Search is not available for this dataset
id
string | submitter
string | authors
string | title
string | comments
string | journal-ref
string | doi
string | report-no
string | categories
string | license
string | abstract
string | versions
list | update_date
timestamp[s] | authors_parsed
sequence | prompt
string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2503.22392 | Xiao Jiang | Xiao Jiang, Grace J. Gang, J. Webster Stayman | Volumetric Material Decomposition Using Spectral Diffusion Posterior
Sampling with a Compressed Polychromatic Forward Model | null | null | null | null | physics.med-ph eess.IV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We have previously introduced Spectral Diffusion Posterior Sampling (Spectral
DPS) as a framework for accurate one-step material decomposition by integrating
analytic spectral system models with priors learned from large datasets. This
work extends the 2D Spectral DPS algorithm to 3D by addressing potentially
limiting large-memory requirements with a pre-trained 2D diffusion model for
slice-by-slice processing and a compressed polychromatic forward model to
ensure accurate physical modeling. Simulation studies demonstrate that the
proposed memory-efficient 3D Spectral DPS enables material decomposition of
clinically significant volume sizes. Quantitative analysis reveals that
Spectral DPS outperforms other deep-learning algorithms, such as InceptNet and
conditional DDPM in contrast quantification, inter-slice continuity, and
resolution preservation. This study establishes a foundation for advancing
one-step material decomposition in volumetric spectral CT.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 12:52:59 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Jiang",
"Xiao",
""
],
[
"Gang",
"Grace J.",
""
],
[
"Stayman",
"J. Webster",
""
]
] | TITLE: Volumetric Material Decomposition Using Spectral Diffusion Posterior
Sampling with a Compressed Polychromatic Forward Model
ABSTRACT: We have previously introduced Spectral Diffusion Posterior Sampling (Spectral
DPS) as a framework for accurate one-step material decomposition by integrating
analytic spectral system models with priors learned from large datasets. This
work extends the 2D Spectral DPS algorithm to 3D by addressing potentially
limiting large-memory requirements with a pre-trained 2D diffusion model for
slice-by-slice processing and a compressed polychromatic forward model to
ensure accurate physical modeling. Simulation studies demonstrate that the
proposed memory-efficient 3D Spectral DPS enables material decomposition of
clinically significant volume sizes. Quantitative analysis reveals that
Spectral DPS outperforms other deep-learning algorithms, such as InceptNet and
conditional DDPM in contrast quantification, inter-slice continuity, and
resolution preservation. This study establishes a foundation for advancing
one-step material decomposition in volumetric spectral CT.
|
2503.22394 | Rulin Zhou | Rulin Zhou and Wenlong He and An Wang and Qiqi Yao and Haijun Hu and
Jiankun Wang and Xi Zhang an Hongliang Ren | Endo-TTAP: Robust Endoscopic Tissue Tracking via Multi-Facet Guided
Attention and Hybrid Flow-point Supervision | null | null | null | null | cs.CV cs.AI | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Accurate tissue point tracking in endoscopic videos is critical for
robotic-assisted surgical navigation and scene understanding, but remains
challenging due to complex deformations, instrument occlusion, and the scarcity
of dense trajectory annotations. Existing methods struggle with long-term
tracking under these conditions due to limited feature utilization and
annotation dependence. We present Endo-TTAP, a novel framework addressing these
challenges through: (1) A Multi-Facet Guided Attention (MFGA) module that
synergizes multi-scale flow dynamics, DINOv2 semantic embeddings, and explicit
motion patterns to jointly predict point positions with uncertainty and
occlusion awareness; (2) A two-stage curriculum learning strategy employing an
Auxiliary Curriculum Adapter (ACA) for progressive initialization and hybrid
supervision. Stage I utilizes synthetic data with optical flow ground truth for
uncertainty-occlusion regularization, while Stage II combines unsupervised flow
consistency and semi-supervised learning with refined pseudo-labels from
off-the-shelf trackers. Extensive validation on two MICCAI Challenge datasets
and our collected dataset demonstrates that Endo-TTAP achieves state-of-the-art
performance in tissue point tracking, particularly in scenarios characterized
by complex endoscopic conditions. The source code and dataset will be available
at https://anonymous.4open.science/r/Endo-TTAP-36E5.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:00:07 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Zhou",
"Rulin",
""
],
[
"He",
"Wenlong",
""
],
[
"Wang",
"An",
""
],
[
"Yao",
"Qiqi",
""
],
[
"Hu",
"Haijun",
""
],
[
"Wang",
"Jiankun",
""
],
[
"Ren",
"Xi Zhang an Hongliang",
""
]
] | TITLE: Endo-TTAP: Robust Endoscopic Tissue Tracking via Multi-Facet Guided
Attention and Hybrid Flow-point Supervision
ABSTRACT: Accurate tissue point tracking in endoscopic videos is critical for
robotic-assisted surgical navigation and scene understanding, but remains
challenging due to complex deformations, instrument occlusion, and the scarcity
of dense trajectory annotations. Existing methods struggle with long-term
tracking under these conditions due to limited feature utilization and
annotation dependence. We present Endo-TTAP, a novel framework addressing these
challenges through: (1) A Multi-Facet Guided Attention (MFGA) module that
synergizes multi-scale flow dynamics, DINOv2 semantic embeddings, and explicit
motion patterns to jointly predict point positions with uncertainty and
occlusion awareness; (2) A two-stage curriculum learning strategy employing an
Auxiliary Curriculum Adapter (ACA) for progressive initialization and hybrid
supervision. Stage I utilizes synthetic data with optical flow ground truth for
uncertainty-occlusion regularization, while Stage II combines unsupervised flow
consistency and semi-supervised learning with refined pseudo-labels from
off-the-shelf trackers. Extensive validation on two MICCAI Challenge datasets
and our collected dataset demonstrates that Endo-TTAP achieves state-of-the-art
performance in tissue point tracking, particularly in scenarios characterized
by complex endoscopic conditions. The source code and dataset will be available
at https://anonymous.4open.science/r/Endo-TTAP-36E5.
|
2503.22395 | Tereza Vrabcov\'a | Tereza Vrabcov\'a, Marek Kadl\v{c}\'ik, Petr Sojka, Michal
\v{S}tef\'anik, Michal Spiegel | Negation: A Pink Elephant in the Large Language Models' Room? | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Negations are key to determining sentence meaning, making them essential for
logical reasoning. Despite their importance, negations pose a substantial
challenge for large language models (LLMs) and remain underexplored.
We construct two multilingual natural language inference (NLI) datasets with
\textit{paired} examples differing in negation. We investigate how model size
and language impact its ability to handle negation correctly by evaluating
popular LLMs.
Contrary to previous work, we show that increasing the model size
consistently improves the models' ability to handle negations. Furthermore, we
find that both the models' reasoning accuracy and robustness to negation are
language-dependent and that the length and explicitness of the premise have a
greater impact on robustness than language.
Our datasets can facilitate further research and improvements of language
model reasoning in multilingual settings.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:04:41 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Vrabcová",
"Tereza",
""
],
[
"Kadlčík",
"Marek",
""
],
[
"Sojka",
"Petr",
""
],
[
"Štefánik",
"Michal",
""
],
[
"Spiegel",
"Michal",
""
]
] | TITLE: Negation: A Pink Elephant in the Large Language Models' Room?
ABSTRACT: Negations are key to determining sentence meaning, making them essential for
logical reasoning. Despite their importance, negations pose a substantial
challenge for large language models (LLMs) and remain underexplored.
We construct two multilingual natural language inference (NLI) datasets with
\textit{paired} examples differing in negation. We investigate how model size
and language impact its ability to handle negation correctly by evaluating
popular LLMs.
Contrary to previous work, we show that increasing the model size
consistently improves the models' ability to handle negations. Furthermore, we
find that both the models' reasoning accuracy and robustness to negation are
language-dependent and that the length and explicitness of the premise have a
greater impact on robustness than language.
Our datasets can facilitate further research and improvements of language
model reasoning in multilingual settings.
|
2503.22397 | Vida Adeli | Vida Adeli, Soroush Mehraban, Majid Mirmehdi, Alan Whone, Benjamin
Filtjens, Amirhossein Dadashzadeh, Alfonso Fasano, Andrea Iaboni Babak Taati | GAITGen: Disentangled Motion-Pathology Impaired Gait Generative Model --
Bringing Motion Generation to the Clinical Domain | null | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Gait analysis is crucial for the diagnosis and monitoring of movement
disorders like Parkinson's Disease. While computer vision models have shown
potential for objectively evaluating parkinsonian gait, their effectiveness is
limited by scarce clinical datasets and the challenge of collecting large and
well-labelled data, impacting model accuracy and risk of bias. To address these
gaps, we propose GAITGen, a novel framework that generates realistic gait
sequences conditioned on specified pathology severity levels. GAITGen employs a
Conditional Residual Vector Quantized Variational Autoencoder to learn
disentangled representations of motion dynamics and pathology-specific factors,
coupled with Mask and Residual Transformers for conditioned sequence
generation. GAITGen generates realistic, diverse gait sequences across severity
levels, enriching datasets and enabling large-scale model training in
parkinsonian gait analysis. Experiments on our new PD-GaM (real) dataset
demonstrate that GAITGen outperforms adapted state-of-the-art models in both
reconstruction fidelity and generation quality, accurately capturing critical
pathology-specific gait features. A clinical user study confirms the realism
and clinical relevance of our generated sequences. Moreover, incorporating
GAITGen-generated data into downstream tasks improves parkinsonian gait
severity estimation, highlighting its potential for advancing clinical gait
analysis.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:06:45 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Adeli",
"Vida",
""
],
[
"Mehraban",
"Soroush",
""
],
[
"Mirmehdi",
"Majid",
""
],
[
"Whone",
"Alan",
""
],
[
"Filtjens",
"Benjamin",
""
],
[
"Dadashzadeh",
"Amirhossein",
""
],
[
"Fasano",
"Alfonso",
""
],
[
"Taati",
"Andrea Iaboni Babak",
""
]
] | TITLE: GAITGen: Disentangled Motion-Pathology Impaired Gait Generative Model --
Bringing Motion Generation to the Clinical Domain
ABSTRACT: Gait analysis is crucial for the diagnosis and monitoring of movement
disorders like Parkinson's Disease. While computer vision models have shown
potential for objectively evaluating parkinsonian gait, their effectiveness is
limited by scarce clinical datasets and the challenge of collecting large and
well-labelled data, impacting model accuracy and risk of bias. To address these
gaps, we propose GAITGen, a novel framework that generates realistic gait
sequences conditioned on specified pathology severity levels. GAITGen employs a
Conditional Residual Vector Quantized Variational Autoencoder to learn
disentangled representations of motion dynamics and pathology-specific factors,
coupled with Mask and Residual Transformers for conditioned sequence
generation. GAITGen generates realistic, diverse gait sequences across severity
levels, enriching datasets and enabling large-scale model training in
parkinsonian gait analysis. Experiments on our new PD-GaM (real) dataset
demonstrate that GAITGen outperforms adapted state-of-the-art models in both
reconstruction fidelity and generation quality, accurately capturing critical
pathology-specific gait features. A clinical user study confirms the realism
and clinical relevance of our generated sequences. Moreover, incorporating
GAITGen-generated data into downstream tasks improves parkinsonian gait
severity estimation, highlighting its potential for advancing clinical gait
analysis.
|
2503.22398 | David Fischinger | David Fischinger and Martin Boyer | DF-Net: The Digital Forensics Network for Image Forgery Detection | Published in 2023 at the 25th Irish Machine Vision and Image
Processing Conference (IMVIP),
https://iprcs.github.io/pdf/IMVIP2023_Proceeding.pdf | 2023 | 25th Irish Machine Vision and Image Processing Conference
(IMVIP) | ISBN: 978-0-9934207-8-8 | 10.5281/zenodo.8214996 10.5281/zenodo.8142658 | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | The orchestrated manipulation of public opinion, particularly through
manipulated images, often spread via online social networks (OSN), has become a
serious threat to society. In this paper we introduce the Digital Forensics Net
(DF-Net), a deep neural network for pixel-wise image forgery detection. The
released model outperforms several state-of-the-art methods on four established
benchmark datasets. Most notably, DF-Net's detection is robust against lossy
image operations (e.g resizing, compression) as they are automatically
performed by social networks.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:06:59 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Fischinger",
"David",
""
],
[
"Boyer",
"Martin",
""
]
] | TITLE: DF-Net: The Digital Forensics Network for Image Forgery Detection
ABSTRACT: The orchestrated manipulation of public opinion, particularly through
manipulated images, often spread via online social networks (OSN), has become a
serious threat to society. In this paper we introduce the Digital Forensics Net
(DF-Net), a deep neural network for pixel-wise image forgery detection. The
released model outperforms several state-of-the-art methods on four established
benchmark datasets. Most notably, DF-Net's detection is robust against lossy
image operations (e.g resizing, compression) as they are automatically
performed by social networks.
|
2503.22408 | Xiaolei Bian | Xiaolei Bian, Changfu Zou, Bj\"orn Fridholm, Christian Sundvall,
Torsten Wik | Smart Sensing Breaks the Accuracy Barrier in Battery State Monitoring | null | null | null | null | eess.SY cs.SY | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Accurate state-of-charge (SOC) estimation is essential for optimizing battery
performance, ensuring safety, and maximizing economic value. Conventional
current and voltage measurements, however, have inherent limitations in fully
inferring the multiphysics-resolved dynamics inside battery cells. This creates
an accuracy barrier that constrains battery usage and reduces
cost-competitiveness and sustainability across industries dependent on battery
technology. In this work, we introduce an integrated sensor framework that
combines novel mechanical, thermal, gas, optical, and electrical sensors with
traditional measurements to break through this barrier. We generate three
unique datasets with eleven measurement types and propose an explainable
machine-learning approach for SOC estimation. This approach renders the
measured signals and the predictive result of machine learning physically
interpretable with respect to battery SOC, offering fundamental insights into
the time-varying importance of different signals. Our experimental results
reveal a marked increase in SOC estimation accuracy--enhanced from 46.1% to
74.5%--compared to conventional methods. This approach not only advances SOC
monitoring precision but also establishes a foundation for monitoring
additional battery states to further improve safety, extend lifespan, and
facilitate fast charging.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:17:58 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Bian",
"Xiaolei",
""
],
[
"Zou",
"Changfu",
""
],
[
"Fridholm",
"Björn",
""
],
[
"Sundvall",
"Christian",
""
],
[
"Wik",
"Torsten",
""
]
] | TITLE: Smart Sensing Breaks the Accuracy Barrier in Battery State Monitoring
ABSTRACT: Accurate state-of-charge (SOC) estimation is essential for optimizing battery
performance, ensuring safety, and maximizing economic value. Conventional
current and voltage measurements, however, have inherent limitations in fully
inferring the multiphysics-resolved dynamics inside battery cells. This creates
an accuracy barrier that constrains battery usage and reduces
cost-competitiveness and sustainability across industries dependent on battery
technology. In this work, we introduce an integrated sensor framework that
combines novel mechanical, thermal, gas, optical, and electrical sensors with
traditional measurements to break through this barrier. We generate three
unique datasets with eleven measurement types and propose an explainable
machine-learning approach for SOC estimation. This approach renders the
measured signals and the predictive result of machine learning physically
interpretable with respect to battery SOC, offering fundamental insights into
the time-varying importance of different signals. Our experimental results
reveal a marked increase in SOC estimation accuracy--enhanced from 46.1% to
74.5%--compared to conventional methods. This approach not only advances SOC
monitoring precision but also establishes a foundation for monitoring
additional battery states to further improve safety, extend lifespan, and
facilitate fast charging.
|
2503.22411 | Petter T\"ornberg | Petter T\"ornberg and Juliana Chueri | Elite Political Discourse has Become More Toxic in Western Countries | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Toxic and uncivil politics is widely seen as a growing threat to democratic
values and governance, yet our understanding of the drivers and evolution of
political incivility remains limited. Leveraging a novel dataset of nearly 18
million Twitter messages from parliamentarians in 17 countries over five years,
this paper systematically investigates whether politics internationally is
becoming more uncivil, and what are the determinants of political incivility.
Our analysis reveals a marked increase in toxic discourse among political
elites, and that it is associated to radical-right parties and parties in
opposition. Toxicity diminished markedly during the early phase of the COVID-19
pandemic and, surprisingly, during election campaigns. Furthermore, our results
indicate that posts relating to ``culture war'' topics, such as migration and
LGBTQ+ rights, are substantially more toxic than debates focused on welfare or
economic issues. These findings underscore a troubling shift in international
democracies toward an erosion of constructive democratic dialogue.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:21:49 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Törnberg",
"Petter",
""
],
[
"Chueri",
"Juliana",
""
]
] | TITLE: Elite Political Discourse has Become More Toxic in Western Countries
ABSTRACT: Toxic and uncivil politics is widely seen as a growing threat to democratic
values and governance, yet our understanding of the drivers and evolution of
political incivility remains limited. Leveraging a novel dataset of nearly 18
million Twitter messages from parliamentarians in 17 countries over five years,
this paper systematically investigates whether politics internationally is
becoming more uncivil, and what are the determinants of political incivility.
Our analysis reveals a marked increase in toxic discourse among political
elites, and that it is associated to radical-right parties and parties in
opposition. Toxicity diminished markedly during the early phase of the COVID-19
pandemic and, surprisingly, during election campaigns. Furthermore, our results
indicate that posts relating to ``culture war'' topics, such as migration and
LGBTQ+ rights, are substantially more toxic than debates focused on welfare or
economic issues. These findings underscore a troubling shift in international
democracies toward an erosion of constructive democratic dialogue.
|
2503.22417 | David Fischinger | David Fischinger and Martin Boyer | DF2023: The Digital Forensics 2023 Dataset for Image Forgery Detection | Published at the 25th Irish Machine Vision and Image Processing
Conference (IMVIP) --- Proceedings:
https://iprcs.github.io/pdf/IMVIP2023_Proceeding.pdf --- Dataset download:
https://zenodo.org/records/7326540/files/DF2023_train.zip
https://zenodo.org/records/7326540/files/DF2023_val.zip Kaggle:
https://www.kaggle.com/datasets/davidfischinger/df2023-digital-forensics-2023-dataset/data | 2023 | 25th Irish Machine Vision and Image Processing Conference
(IMVIP) | ISBN: 978-0-9934207-8-8 | 10.5281/zenodo.8215043 10.5281/zenodo.7326540 | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | The deliberate manipulation of public opinion, especially through altered
images, which are frequently disseminated through online social networks, poses
a significant danger to society. To fight this issue on a technical level we
support the research community by releasing the Digital Forensics 2023 (DF2023)
training and validation dataset, comprising one million images from four major
forgery categories: splicing, copy-move, enhancement and removal. This dataset
enables an objective comparison of network architectures and can significantly
reduce the time and effort of researchers preparing datasets.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:31:19 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Fischinger",
"David",
""
],
[
"Boyer",
"Martin",
""
]
] | TITLE: DF2023: The Digital Forensics 2023 Dataset for Image Forgery Detection
ABSTRACT: The deliberate manipulation of public opinion, especially through altered
images, which are frequently disseminated through online social networks, poses
a significant danger to society. To fight this issue on a technical level we
support the research community by releasing the Digital Forensics 2023 (DF2023)
training and validation dataset, comprising one million images from four major
forgery categories: splicing, copy-move, enhancement and removal. This dataset
enables an objective comparison of network architectures and can significantly
reduce the time and effort of researchers preparing datasets.
|
2503.22427 | Rajkumar Muthusamy DSc (Tech) | Abhinav Pathak and Rajkumar Muthusamy | Collapse and Collision Aware Grasping for Cluttered Shelf Picking | null | null | null | null | cs.RO | http://creativecommons.org/licenses/by/4.0/ | Efficient and safe retrieval of stacked objects in warehouse environments is
a significant challenge due to complex spatial dependencies and structural
inter-dependencies. Traditional vision-based methods excel at object
localization but often lack the physical reasoning required to predict the
consequences of extraction, leading to unintended collisions and collapses.
This paper proposes a collapse and collision aware grasp planner that
integrates dynamic physics simulations for robotic decision-making. Using a
single image and depth map, an approximate 3D representation of the scene is
reconstructed in a simulation environment, enabling the robot to evaluate
different retrieval strategies before execution. Two approaches 1)
heuristic-based and 2) physics-based are proposed for both single-box
extraction and shelf clearance tasks. Extensive real-world experiments on
structured and unstructured box stacks, along with validation using datasets
from existing databases, show that our physics-aware method significantly
improves efficiency and success rates compared to baseline heuristics.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:42:54 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Pathak",
"Abhinav",
""
],
[
"Muthusamy",
"Rajkumar",
""
]
] | TITLE: Collapse and Collision Aware Grasping for Cluttered Shelf Picking
ABSTRACT: Efficient and safe retrieval of stacked objects in warehouse environments is
a significant challenge due to complex spatial dependencies and structural
inter-dependencies. Traditional vision-based methods excel at object
localization but often lack the physical reasoning required to predict the
consequences of extraction, leading to unintended collisions and collapses.
This paper proposes a collapse and collision aware grasp planner that
integrates dynamic physics simulations for robotic decision-making. Using a
single image and depth map, an approximate 3D representation of the scene is
reconstructed in a simulation environment, enabling the robot to evaluate
different retrieval strategies before execution. Two approaches 1)
heuristic-based and 2) physics-based are proposed for both single-box
extraction and shelf clearance tasks. Extensive real-world experiments on
structured and unstructured box stacks, along with validation using datasets
from existing databases, show that our physics-aware method significantly
improves efficiency and success rates compared to baseline heuristics.
|
2503.22436 | Fuhao Li | Fuhao Li, Huan Jin, Bin Gao, Liaoyuan Fan, Lihui Jiang, Long Zeng | NuGrounding: A Multi-View 3D Visual Grounding Framework in Autonomous
Driving | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Multi-view 3D visual grounding is critical for autonomous driving vehicles to
interpret natural languages and localize target objects in complex
environments. However, existing datasets and methods suffer from coarse-grained
language instructions, and inadequate integration of 3D geometric reasoning
with linguistic comprehension. To this end, we introduce NuGrounding, the first
large-scale benchmark for multi-view 3D visual grounding in autonomous driving.
We present a Hierarchy of Grounding (HoG) method to construct NuGrounding to
generate hierarchical multi-level instructions, ensuring comprehensive coverage
of human instruction patterns. To tackle this challenging dataset, we propose a
novel paradigm that seamlessly combines instruction comprehension abilities of
multi-modal LLMs (MLLMs) with precise localization abilities of specialist
detection models. Our approach introduces two decoupled task tokens and a
context query to aggregate 3D geometric information and semantic instructions,
followed by a fusion decoder to refine spatial-semantic feature fusion for
precise localization. Extensive experiments demonstrate that our method
significantly outperforms the baselines adapted from representative 3D scene
understanding methods by a significant margin and achieves 0.59 in precision
and 0.64 in recall, with improvements of 50.8% and 54.7%.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:55:16 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Li",
"Fuhao",
""
],
[
"Jin",
"Huan",
""
],
[
"Gao",
"Bin",
""
],
[
"Fan",
"Liaoyuan",
""
],
[
"Jiang",
"Lihui",
""
],
[
"Zeng",
"Long",
""
]
] | TITLE: NuGrounding: A Multi-View 3D Visual Grounding Framework in Autonomous
Driving
ABSTRACT: Multi-view 3D visual grounding is critical for autonomous driving vehicles to
interpret natural languages and localize target objects in complex
environments. However, existing datasets and methods suffer from coarse-grained
language instructions, and inadequate integration of 3D geometric reasoning
with linguistic comprehension. To this end, we introduce NuGrounding, the first
large-scale benchmark for multi-view 3D visual grounding in autonomous driving.
We present a Hierarchy of Grounding (HoG) method to construct NuGrounding to
generate hierarchical multi-level instructions, ensuring comprehensive coverage
of human instruction patterns. To tackle this challenging dataset, we propose a
novel paradigm that seamlessly combines instruction comprehension abilities of
multi-modal LLMs (MLLMs) with precise localization abilities of specialist
detection models. Our approach introduces two decoupled task tokens and a
context query to aggregate 3D geometric information and semantic instructions,
followed by a fusion decoder to refine spatial-semantic feature fusion for
precise localization. Extensive experiments demonstrate that our method
significantly outperforms the baselines adapted from representative 3D scene
understanding methods by a significant margin and achieves 0.59 in precision
and 0.64 in recall, with improvements of 50.8% and 54.7%.
|
2503.22437 | Xu Wang Mr | Xu Wang, Shuai Zhang, Baoru Huang, Danail Stoyanov, Evangelos B.
Mazomenos | EndoLRMGS: Complete Endoscopic Scene Reconstruction combining Large
Reconstruction Modelling and Gaussian Splatting | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Complete reconstruction of surgical scenes is crucial for robot-assisted
surgery (RAS). Deep depth estimation is promising but existing works struggle
with depth discontinuities, resulting in noisy predictions at object boundaries
and do not achieve complete reconstruction omitting occluded surfaces. To
address these issues we propose EndoLRMGS, that combines Large Reconstruction
Modelling (LRM) and Gaussian Splatting (GS), for complete surgical scene
reconstruction. GS reconstructs deformable tissues and LRM generates 3D models
for surgical tools while position and scale are subsequently optimized by
introducing orthogonal perspective joint projection optimization (OPjPO) to
enhance accuracy. In experiments on four surgical videos from three public
datasets, our method improves the Intersection-over-union (IoU) of tool 3D
models in 2D projections by>40%. Additionally, EndoLRMGS improves the PSNR of
the tools projection from 3.82% to 11.07%. Tissue rendering quality also
improves, with PSNR increasing from 0.46% to 49.87%, and SSIM from 1.53% to
29.21% across all test videos.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 13:57:12 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Wang",
"Xu",
""
],
[
"Zhang",
"Shuai",
""
],
[
"Huang",
"Baoru",
""
],
[
"Stoyanov",
"Danail",
""
],
[
"Mazomenos",
"Evangelos B.",
""
]
] | TITLE: EndoLRMGS: Complete Endoscopic Scene Reconstruction combining Large
Reconstruction Modelling and Gaussian Splatting
ABSTRACT: Complete reconstruction of surgical scenes is crucial for robot-assisted
surgery (RAS). Deep depth estimation is promising but existing works struggle
with depth discontinuities, resulting in noisy predictions at object boundaries
and do not achieve complete reconstruction omitting occluded surfaces. To
address these issues we propose EndoLRMGS, that combines Large Reconstruction
Modelling (LRM) and Gaussian Splatting (GS), for complete surgical scene
reconstruction. GS reconstructs deformable tissues and LRM generates 3D models
for surgical tools while position and scale are subsequently optimized by
introducing orthogonal perspective joint projection optimization (OPjPO) to
enhance accuracy. In experiments on four surgical videos from three public
datasets, our method improves the Intersection-over-union (IoU) of tool 3D
models in 2D projections by>40%. Additionally, EndoLRMGS improves the PSNR of
the tools projection from 3.82% to 11.07%. Tissue rendering quality also
improves, with PSNR increasing from 0.46% to 49.87%, and SSIM from 1.53% to
29.21% across all test videos.
|
2503.22448 | Graciana Puentes | Graciana Puentes | Comparison between neural network clustering, hierarchical clustering
and k-means clustering: Applications using fluidic lenses | 19 pages, 9 figures | null | null | null | physics.optics cs.LG | http://creativecommons.org/licenses/by/4.0/ | A comparison between neural network clustering (NNC), hierarchical clustering
(HC) and K-means clustering (KMC) is performed to evaluate the computational
superiority of these three machine learning (ML) techniques for organizing
large datasets into clusters. For NNC, a self-organizing map (SOM) training was
applied to a collection of wavefront sensor reconstructions, decomposed in
terms of 15 Zernike coefficients, characterizing the optical aberrations of the
phase front transmitted by fluidic lenses. In order to understand the
distribution and structure of the 15 Zernike variables within an input space,
SOM-neighboring weight distances, SOM-sample hits, SOM-weight positions and
SOM-weight planes were analyzed to form a visual interpretation of the system's
structural properties. In the case of HC, the data was partitioned using a
combined dissimilarity-linkage matrix computation. The effectiveness of this
method was confirmed by a high cophenetic correlation coefficient value
(c=0.9651). Additionally, a maximum number of clusters was established by
setting an inconsistency cutoff of 0.8, yielding a total of 7 clusters for
system segmentation. In addition, a KMC approach was employed to establish a
quantitative measure of clustering segmentation efficiency, obtaining a
sillhoute average value of 0.905 for data segmentation into K=5 non-overlapping
clusters. On the other hand, the NNC analysis revealed that the 15 variables
could be characterized through the collective influence of 8 clusters. It was
established that the formation of clusters through the combined linkage and
dissimilarity algorithms of HC alongside KMC is a more dependable clustering
solution than separate assessment via NNC or HC, where altering the SOM size or
inconsistency cutoff can lead to completely new clustering configurations.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 14:01:12 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Puentes",
"Graciana",
""
]
] | TITLE: Comparison between neural network clustering, hierarchical clustering
and k-means clustering: Applications using fluidic lenses
ABSTRACT: A comparison between neural network clustering (NNC), hierarchical clustering
(HC) and K-means clustering (KMC) is performed to evaluate the computational
superiority of these three machine learning (ML) techniques for organizing
large datasets into clusters. For NNC, a self-organizing map (SOM) training was
applied to a collection of wavefront sensor reconstructions, decomposed in
terms of 15 Zernike coefficients, characterizing the optical aberrations of the
phase front transmitted by fluidic lenses. In order to understand the
distribution and structure of the 15 Zernike variables within an input space,
SOM-neighboring weight distances, SOM-sample hits, SOM-weight positions and
SOM-weight planes were analyzed to form a visual interpretation of the system's
structural properties. In the case of HC, the data was partitioned using a
combined dissimilarity-linkage matrix computation. The effectiveness of this
method was confirmed by a high cophenetic correlation coefficient value
(c=0.9651). Additionally, a maximum number of clusters was established by
setting an inconsistency cutoff of 0.8, yielding a total of 7 clusters for
system segmentation. In addition, a KMC approach was employed to establish a
quantitative measure of clustering segmentation efficiency, obtaining a
sillhoute average value of 0.905 for data segmentation into K=5 non-overlapping
clusters. On the other hand, the NNC analysis revealed that the 15 variables
could be characterized through the collective influence of 8 clusters. It was
established that the formation of clusters through the combined linkage and
dissimilarity algorithms of HC alongside KMC is a more dependable clustering
solution than separate assessment via NNC or HC, where altering the SOM size or
inconsistency cutoff can lead to completely new clustering configurations.
|
2503.22454 | Ayan Majumdar | Ayan Majumdar and Deborah D. Kanubala and Kavya Gupta and Isabel
Valera | A Causal Framework to Measure and Mitigate Non-binary Treatment
Discrimination | 24 pages, 5 figures | null | null | null | cs.LG cs.AI | http://creativecommons.org/licenses/by/4.0/ | Fairness studies of algorithmic decision-making systems often simplify
complex decision processes, such as bail or loan approvals, into binary
classification tasks. However, these approaches overlook that such decisions
are not inherently binary (e.g., approve or not approve bail or loan); they
also involve non-binary treatment decisions (e.g., bail conditions or loan
terms) that can influence the downstream outcomes (e.g., loan repayment or
reoffending). In this paper, we argue that non-binary treatment decisions are
integral to the decision process and controlled by decision-makers and,
therefore, should be central to fairness analyses in algorithmic
decision-making. We propose a causal framework that extends fairness analyses
and explicitly distinguishes between decision-subjects' covariates and the
treatment decisions. This specification allows decision-makers to use our
framework to (i) measure treatment disparity and its downstream effects in
historical data and, using counterfactual reasoning, (ii) mitigate the impact
of past unfair treatment decisions when automating decision-making. We use our
framework to empirically analyze four widely used loan approval datasets to
reveal potential disparity in non-binary treatment decisions and their
discriminatory impact on outcomes, highlighting the need to incorporate
treatment decisions in fairness assessments. Moreover, by intervening in
treatment decisions, we show that our framework effectively mitigates treatment
discrimination from historical data to ensure fair risk score estimation and
(non-binary) decision-making processes that benefit all stakeholders.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 14:06:35 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Majumdar",
"Ayan",
""
],
[
"Kanubala",
"Deborah D.",
""
],
[
"Gupta",
"Kavya",
""
],
[
"Valera",
"Isabel",
""
]
] | TITLE: A Causal Framework to Measure and Mitigate Non-binary Treatment
Discrimination
ABSTRACT: Fairness studies of algorithmic decision-making systems often simplify
complex decision processes, such as bail or loan approvals, into binary
classification tasks. However, these approaches overlook that such decisions
are not inherently binary (e.g., approve or not approve bail or loan); they
also involve non-binary treatment decisions (e.g., bail conditions or loan
terms) that can influence the downstream outcomes (e.g., loan repayment or
reoffending). In this paper, we argue that non-binary treatment decisions are
integral to the decision process and controlled by decision-makers and,
therefore, should be central to fairness analyses in algorithmic
decision-making. We propose a causal framework that extends fairness analyses
and explicitly distinguishes between decision-subjects' covariates and the
treatment decisions. This specification allows decision-makers to use our
framework to (i) measure treatment disparity and its downstream effects in
historical data and, using counterfactual reasoning, (ii) mitigate the impact
of past unfair treatment decisions when automating decision-making. We use our
framework to empirically analyze four widely used loan approval datasets to
reveal potential disparity in non-binary treatment decisions and their
discriminatory impact on outcomes, highlighting the need to incorporate
treatment decisions in fairness assessments. Moreover, by intervening in
treatment decisions, we show that our framework effectively mitigates treatment
discrimination from historical data to ensure fair risk score estimation and
(non-binary) decision-making processes that benefit all stakeholders.
|
2503.22460 | Xin He | YongKang Yan, Zeqian Gan, Luying Hu, Xinrui Xu, Ran Kang, Chengwei
Qian, Jianqiang Mei, Paul Beckett, William Shieh, Rui Yin, Xin He, Xu Liu | High-Dimensional Encoding Computational Imaging | 18 pages, 10 figures, 1 table | null | null | null | physics.optics | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | High-dimensional imaging technology has demonstrated significant research
value across diverse fields, including environmental monitoring, agricultural
inspection, and biomedical imaging, through integrating spatial (X*Y),
spectral, and polarization detection functionalities. Here, we report a
High-Dimensional encoding computational imaging technique, utilizing 4
high-dimensional encoders (HDE1-4) and a high-dimensional neural network (HDNN)
to reconstruct 80 high-dimensional images of the target. The system efficiently
acquires spectral-polarization information, spanning a wavelength range of
400-800 nm at intervals of 20 nm, obtaining 20 spectral datasets. Each dataset
contains images captured at 4 polarization angles (0{\deg}, 45{\deg}, 90{\deg},
and -45{\deg}), and the image resolution can reach up to 1280 * 960 pixels.
Achieving a reconstruction ratio 1:20. Experimental validation confirms that
the spectral reconstruction error consistently remains below 0.14%. Extensive
high-dimensional imaging experiments were conducted under indoor and outdoor
conditions, showing the system's significant adaptability and robustness in
various environments. Compared to traditional imaging devices, such as
hyperspectral cameras that could only acquire spectral information, while
polarization cameras are limited to polarization imaging, this integrated
system successfully overcomes these technological constraints, providing an
innovative and efficient solution for high-dimensional optical sensing
applications.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 14:13:32 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Yan",
"YongKang",
""
],
[
"Gan",
"Zeqian",
""
],
[
"Hu",
"Luying",
""
],
[
"Xu",
"Xinrui",
""
],
[
"Kang",
"Ran",
""
],
[
"Qian",
"Chengwei",
""
],
[
"Mei",
"Jianqiang",
""
],
[
"Beckett",
"Paul",
""
],
[
"Shieh",
"William",
""
],
[
"Yin",
"Rui",
""
],
[
"He",
"Xin",
""
],
[
"Liu",
"Xu",
""
]
] | TITLE: High-Dimensional Encoding Computational Imaging
ABSTRACT: High-dimensional imaging technology has demonstrated significant research
value across diverse fields, including environmental monitoring, agricultural
inspection, and biomedical imaging, through integrating spatial (X*Y),
spectral, and polarization detection functionalities. Here, we report a
High-Dimensional encoding computational imaging technique, utilizing 4
high-dimensional encoders (HDE1-4) and a high-dimensional neural network (HDNN)
to reconstruct 80 high-dimensional images of the target. The system efficiently
acquires spectral-polarization information, spanning a wavelength range of
400-800 nm at intervals of 20 nm, obtaining 20 spectral datasets. Each dataset
contains images captured at 4 polarization angles (0{\deg}, 45{\deg}, 90{\deg},
and -45{\deg}), and the image resolution can reach up to 1280 * 960 pixels.
Achieving a reconstruction ratio 1:20. Experimental validation confirms that
the spectral reconstruction error consistently remains below 0.14%. Extensive
high-dimensional imaging experiments were conducted under indoor and outdoor
conditions, showing the system's significant adaptability and robustness in
various environments. Compared to traditional imaging devices, such as
hyperspectral cameras that could only acquire spectral information, while
polarization cameras are limited to polarization imaging, this integrated
system successfully overcomes these technological constraints, providing an
innovative and efficient solution for high-dimensional optical sensing
applications.
|
2503.22462 | Krispin Wandel | Krispin Wandel, Hesheng Wang | SemAlign3D: Semantic Correspondence between RGB-Images through Aligning
3D Object-Class Representations | Accepted to CVPR 2025. Poster:
https://cvpr.thecvf.com/virtual/2025/poster/32799 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Semantic correspondence made tremendous progress through the recent
advancements of large vision models (LVM). While these LVMs have been shown to
reliably capture local semantics, the same can currently not be said for
capturing global geometric relationships between semantic object regions. This
problem leads to unreliable performance for semantic correspondence between
images with extreme view variation. In this work, we aim to leverage monocular
depth estimates to capture these geometric relationships for more robust and
data-efficient semantic correspondence. First, we introduce a simple but
effective method to build 3D object-class representations from monocular depth
estimates and LVM features using a sparsely annotated image correspondence
dataset. Second, we formulate an alignment energy that can be minimized using
gradient descent to obtain an alignment between the 3D object-class
representation and the object-class instance in the input RGB-image. Our method
achieves state-of-the-art matching accuracy in multiple categories on the
challenging SPair-71k dataset, increasing the [email protected] score by more than 10
points on three categories and overall by 3.3 points from 85.6% to 88.9%.
Additional resources and code are available at https://dub.sh/semalign3d.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 14:14:19 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Wandel",
"Krispin",
""
],
[
"Wang",
"Hesheng",
""
]
] | TITLE: SemAlign3D: Semantic Correspondence between RGB-Images through Aligning
3D Object-Class Representations
ABSTRACT: Semantic correspondence made tremendous progress through the recent
advancements of large vision models (LVM). While these LVMs have been shown to
reliably capture local semantics, the same can currently not be said for
capturing global geometric relationships between semantic object regions. This
problem leads to unreliable performance for semantic correspondence between
images with extreme view variation. In this work, we aim to leverage monocular
depth estimates to capture these geometric relationships for more robust and
data-efficient semantic correspondence. First, we introduce a simple but
effective method to build 3D object-class representations from monocular depth
estimates and LVM features using a sparsely annotated image correspondence
dataset. Second, we formulate an alignment energy that can be minimized using
gradient descent to obtain an alignment between the 3D object-class
representation and the object-class instance in the input RGB-image. Our method
achieves state-of-the-art matching accuracy in multiple categories on the
challenging SPair-71k dataset, increasing the [email protected] score by more than 10
points on three categories and overall by 3.3 points from 85.6% to 88.9%.
Additional resources and code are available at https://dub.sh/semalign3d.
|
2503.22473 | Hanchao Liu | Hanchao Liu, Rongjun Li, Weimin Xiong, Ziyu Zhou, Wei Peng | WorkTeam: Constructing Workflows from Natural Language with Multi-Agents | Accepted in NAACL 2025 Industry Track | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Workflows play a crucial role in enhancing enterprise efficiency by
orchestrating complex processes with multiple tools or components. However,
hand-crafted workflow construction requires expert knowledge, presenting
significant technical barriers. Recent advancements in Large Language Models
(LLMs) have improved the generation of workflows from natural language
instructions (aka NL2Workflow), yet existing single LLM agent-based methods
face performance degradation on complex tasks due to the need for specialized
knowledge and the strain of task-switching. To tackle these challenges, we
propose WorkTeam, a multi-agent NL2Workflow framework comprising a supervisor,
orchestrator, and filler agent, each with distinct roles that collaboratively
enhance the conversion process. As there are currently no publicly available
NL2Workflow benchmarks, we also introduce the HW-NL2Workflow dataset, which
includes 3,695 real-world business samples for training and evaluation.
Experimental results show that our approach significantly increases the success
rate of workflow construction, providing a novel and effective solution for
enterprise NL2Workflow services.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 14:33:29 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Liu",
"Hanchao",
""
],
[
"Li",
"Rongjun",
""
],
[
"Xiong",
"Weimin",
""
],
[
"Zhou",
"Ziyu",
""
],
[
"Peng",
"Wei",
""
]
] | TITLE: WorkTeam: Constructing Workflows from Natural Language with Multi-Agents
ABSTRACT: Workflows play a crucial role in enhancing enterprise efficiency by
orchestrating complex processes with multiple tools or components. However,
hand-crafted workflow construction requires expert knowledge, presenting
significant technical barriers. Recent advancements in Large Language Models
(LLMs) have improved the generation of workflows from natural language
instructions (aka NL2Workflow), yet existing single LLM agent-based methods
face performance degradation on complex tasks due to the need for specialized
knowledge and the strain of task-switching. To tackle these challenges, we
propose WorkTeam, a multi-agent NL2Workflow framework comprising a supervisor,
orchestrator, and filler agent, each with distinct roles that collaboratively
enhance the conversion process. As there are currently no publicly available
NL2Workflow benchmarks, we also introduce the HW-NL2Workflow dataset, which
includes 3,695 real-world business samples for training and evaluation.
Experimental results show that our approach significantly increases the success
rate of workflow construction, providing a novel and effective solution for
enterprise NL2Workflow services.
|
2503.22475 | Bo Shen | Chenyang Li, Tanmay Sunil Kapure, Prokash Chandra Roy, Zhengtao Gan,
Bo Shen | DeepOFormer: Deep Operator Learning with Domain-informed Features for
Fatigue Life Prediction | 6 pages, 4 figures | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Fatigue life characterizes the duration a material can function before
failure under specific environmental conditions, and is traditionally assessed
using stress-life (S-N) curves. While machine learning and deep learning offer
promising results for fatigue life prediction, they face the overfitting
challenge because of the small size of fatigue experimental data in specific
materials. To address this challenge, we propose, DeepOFormer, by formulating
S-N curve prediction as an operator learning problem. DeepOFormer improves the
deep operator learning framework with a transformer-based encoder and a mean L2
relative error loss function. We also consider Stussi, Weibull, and Pascual and
Meeker (PM) features as domain-informed features. These features are motivated
by empirical fatigue models. To evaluate the performance of our DeepOFormer, we
compare it with different deep learning models and XGBoost on a dataset with 54
S-N curves of aluminum alloys. With seven different aluminum alloys selected
for testing, our DeepOFormer achieves an R2 of 0.9515, a mean absolute error of
0.2080, and a mean relative error of 0.5077, significantly outperforming
state-of-the-art deep/machine learning methods including DeepONet,
TabTransformer, and XGBoost, etc. The results highlight that our Deep0Former
integrating with domain-informed features substantially improves prediction
accuracy and generalization capabilities for fatigue life prediction in
aluminum alloys.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 14:34:35 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Li",
"Chenyang",
""
],
[
"Kapure",
"Tanmay Sunil",
""
],
[
"Roy",
"Prokash Chandra",
""
],
[
"Gan",
"Zhengtao",
""
],
[
"Shen",
"Bo",
""
]
] | TITLE: DeepOFormer: Deep Operator Learning with Domain-informed Features for
Fatigue Life Prediction
ABSTRACT: Fatigue life characterizes the duration a material can function before
failure under specific environmental conditions, and is traditionally assessed
using stress-life (S-N) curves. While machine learning and deep learning offer
promising results for fatigue life prediction, they face the overfitting
challenge because of the small size of fatigue experimental data in specific
materials. To address this challenge, we propose, DeepOFormer, by formulating
S-N curve prediction as an operator learning problem. DeepOFormer improves the
deep operator learning framework with a transformer-based encoder and a mean L2
relative error loss function. We also consider Stussi, Weibull, and Pascual and
Meeker (PM) features as domain-informed features. These features are motivated
by empirical fatigue models. To evaluate the performance of our DeepOFormer, we
compare it with different deep learning models and XGBoost on a dataset with 54
S-N curves of aluminum alloys. With seven different aluminum alloys selected
for testing, our DeepOFormer achieves an R2 of 0.9515, a mean absolute error of
0.2080, and a mean relative error of 0.5077, significantly outperforming
state-of-the-art deep/machine learning methods including DeepONet,
TabTransformer, and XGBoost, etc. The results highlight that our Deep0Former
integrating with domain-informed features substantially improves prediction
accuracy and generalization capabilities for fatigue life prediction in
aluminum alloys.
|
2503.22498 | Jing Li | Jing Li and Hao Sun | Learnable cut flow | 26 pages, 33 figures | null | null | null | cs.LG hep-ph | http://creativecommons.org/licenses/by/4.0/ | Neural networks have emerged as a powerful paradigm for tasks in high energy
physics, yet their opaque training process renders them as a black box. In
contrast, the traditional cut flow method offers simplicity and
interpretability but demands human effort to identify optimal boundaries. To
merge the strengths of both approaches, we propose the Learnable Cut Flow
(LCF), a neural network that transforms the traditional cut selection into a
fully differentiable, data-driven process. LCF implements two cut
strategies-parallel, where observable distributions are treated independently,
and sequential, where prior cuts shape subsequent ones-to flexibly determine
optimal boundaries. Building on this, we introduce the Learnable Importance, a
metric that quantifies feature importance and adjusts their contributions to
the loss accordingly, offering model-driven insights unlike ad-hoc metrics. To
ensure differentiability, a modified loss function replaces hard cuts with mask
operations, preserving data shape throughout the training process. LCF is
tested on six varied mock datasets and a realistic diboson vs. QCD dataset.
Results demonstrate that LCF (1) accurately learns cut boundaries across
typical feature distributions in both parallel and sequential strategies, (2)
assigns higher importance to discriminative features with minimal overlap, (3)
handles redundant or correlated features robustly, and (4) performs effectively
in real-world scenarios. In diboson dataset, LCF initially underperforms
boosted decision trees and multiplayer perceptrons when using all observables.
However, pruning less critical features-guided by learned importance-boosts its
performance to match or exceed these baselines. LCF bridges the gap between
traditional cut flow method and modern black-box neural networks, delivering
actionable insights into the training process and feature importance.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 15:04:06 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Li",
"Jing",
""
],
[
"Sun",
"Hao",
""
]
] | TITLE: Learnable cut flow
ABSTRACT: Neural networks have emerged as a powerful paradigm for tasks in high energy
physics, yet their opaque training process renders them as a black box. In
contrast, the traditional cut flow method offers simplicity and
interpretability but demands human effort to identify optimal boundaries. To
merge the strengths of both approaches, we propose the Learnable Cut Flow
(LCF), a neural network that transforms the traditional cut selection into a
fully differentiable, data-driven process. LCF implements two cut
strategies-parallel, where observable distributions are treated independently,
and sequential, where prior cuts shape subsequent ones-to flexibly determine
optimal boundaries. Building on this, we introduce the Learnable Importance, a
metric that quantifies feature importance and adjusts their contributions to
the loss accordingly, offering model-driven insights unlike ad-hoc metrics. To
ensure differentiability, a modified loss function replaces hard cuts with mask
operations, preserving data shape throughout the training process. LCF is
tested on six varied mock datasets and a realistic diboson vs. QCD dataset.
Results demonstrate that LCF (1) accurately learns cut boundaries across
typical feature distributions in both parallel and sequential strategies, (2)
assigns higher importance to discriminative features with minimal overlap, (3)
handles redundant or correlated features robustly, and (4) performs effectively
in real-world scenarios. In diboson dataset, LCF initially underperforms
boosted decision trees and multiplayer perceptrons when using all observables.
However, pruning less critical features-guided by learned importance-boosts its
performance to match or exceed these baselines. LCF bridges the gap between
traditional cut flow method and modern black-box neural networks, delivering
actionable insights into the training process and feature importance.
|
2503.22510 | Yongmin Li | Simran Kaur Ghatoray and Yongmin Li | Automated UX Insights from User Research Videos by Integrating Facial
Emotion and Text Sentiment | null | null | null | null | cs.HC | http://creativecommons.org/licenses/by/4.0/ | Emotion recognition technology has been studied from the past decade. With
its growing importance and applications such as customer service, medical,
education, etc., this research study aims to explore its potential and
importance in the field of User experience evaluation. Recognizing and keeping
track of user emotions in user research video is important to understand user
needs and expectations from a service/product. Little research has been done
that focuses on automating emotion extraction from a video where more than one
modality has been incorporated in the field of UX. The study aims at
implementing different modalities such as facial emotion recognition,
speech-to-text and text-based emotion recognition for capturing emotional
nuances from a user research video and extract meaningful actionable insights.
For selection of facial emotion recognition model, 10 pre-trained models were
evaluated on three benchmark datasets i.e. FER-2013, AffectNet and CK+,
selecting the model with most generalization ability. To extract speech and
convert to text, OpenAI's Whisper model was implemented and finally the
emotions from text were recognized using a pre-trained model available at
HuggingFace website having an evaluation accuracy more than 95%. The study also
integrates the gathered data using temporal alignment and fusion for deeper and
contextual insights. The study further demonstrates a way of automating data
analysis through PandasAI Python library where OpenAI's GPT-4o model was
implemented along with a discussion on other possible solutions. This study is
an attempt to demonstrate a proof of concept where automated meaningful
insights are extracted from a video based on user emotions.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 15:14:08 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Ghatoray",
"Simran Kaur",
""
],
[
"Li",
"Yongmin",
""
]
] | TITLE: Automated UX Insights from User Research Videos by Integrating Facial
Emotion and Text Sentiment
ABSTRACT: Emotion recognition technology has been studied from the past decade. With
its growing importance and applications such as customer service, medical,
education, etc., this research study aims to explore its potential and
importance in the field of User experience evaluation. Recognizing and keeping
track of user emotions in user research video is important to understand user
needs and expectations from a service/product. Little research has been done
that focuses on automating emotion extraction from a video where more than one
modality has been incorporated in the field of UX. The study aims at
implementing different modalities such as facial emotion recognition,
speech-to-text and text-based emotion recognition for capturing emotional
nuances from a user research video and extract meaningful actionable insights.
For selection of facial emotion recognition model, 10 pre-trained models were
evaluated on three benchmark datasets i.e. FER-2013, AffectNet and CK+,
selecting the model with most generalization ability. To extract speech and
convert to text, OpenAI's Whisper model was implemented and finally the
emotions from text were recognized using a pre-trained model available at
HuggingFace website having an evaluation accuracy more than 95%. The study also
integrates the gathered data using temporal alignment and fusion for deeper and
contextual insights. The study further demonstrates a way of automating data
analysis through PandasAI Python library where OpenAI's GPT-4o model was
implemented along with a discussion on other possible solutions. This study is
an attempt to demonstrate a proof of concept where automated meaningful
insights are extracted from a video based on user emotions.
|
2503.22513 | Martin Ki\v{s}\v{s} | Martin Ki\v{s}\v{s} and Michal Hradi\v{s} | Masked Self-Supervised Pre-Training for Text Recognition Transformers on
Large-Scale Datasets | 18 pages, 7 tables, 6 figures; Submitted to ICDAR25 | null | null | null | cs.CV cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Self-supervised learning has emerged as a powerful approach for leveraging
large-scale unlabeled data to improve model performance in various domains. In
this paper, we explore masked self-supervised pre-training for text recognition
transformers. Specifically, we propose two modifications to the pre-training
phase: progressively increasing the masking probability, and modifying the loss
function to incorporate both masked and non-masked patches. We conduct
extensive experiments using a dataset of 50M unlabeled text lines for
pre-training and four differently sized annotated datasets for fine-tuning.
Furthermore, we compare our pre-trained models against those trained with
transfer learning, demonstrating the effectiveness of the self-supervised
pre-training. In particular, pre-training consistently improves the character
error rate of models, in some cases up to 30 % relatively. It is also on par
with transfer learning but without relying on extra annotated text lines.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 15:16:48 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Kišš",
"Martin",
""
],
[
"Hradiš",
"Michal",
""
]
] | TITLE: Masked Self-Supervised Pre-Training for Text Recognition Transformers on
Large-Scale Datasets
ABSTRACT: Self-supervised learning has emerged as a powerful approach for leveraging
large-scale unlabeled data to improve model performance in various domains. In
this paper, we explore masked self-supervised pre-training for text recognition
transformers. Specifically, we propose two modifications to the pre-training
phase: progressively increasing the masking probability, and modifying the loss
function to incorporate both masked and non-masked patches. We conduct
extensive experiments using a dataset of 50M unlabeled text lines for
pre-training and four differently sized annotated datasets for fine-tuning.
Furthermore, we compare our pre-trained models against those trained with
transfer learning, demonstrating the effectiveness of the self-supervised
pre-training. In particular, pre-training consistently improves the character
error rate of models, in some cases up to 30 % relatively. It is also on par
with transfer learning but without relying on extra annotated text lines.
|
2503.22524 | Shuze Wang | Shuze Wang, Yunpeng Mei, Hongjie Cao, Yetian Yuan, Gang Wang, Jian
Sun, Jie Chen | Robust Offline Imitation Learning Through State-level Trajectory
Stitching | null | null | null | null | cs.RO cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Imitation learning (IL) has proven effective for enabling robots to acquire
visuomotor skills through expert demonstrations. However, traditional IL
methods are limited by their reliance on high-quality, often scarce, expert
data, and suffer from covariate shift. To address these challenges, recent
advances in offline IL have incorporated suboptimal, unlabeled datasets into
the training. In this paper, we propose a novel approach to enhance policy
learning from mixed-quality offline datasets by leveraging task-relevant
trajectory fragments and rich environmental dynamics. Specifically, we
introduce a state-based search framework that stitches state-action pairs from
imperfect demonstrations, generating more diverse and informative training
trajectories. Experimental results on standard IL benchmarks and real-world
robotic tasks showcase that our proposed method significantly improves both
generalization and performance.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 15:28:36 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Wang",
"Shuze",
""
],
[
"Mei",
"Yunpeng",
""
],
[
"Cao",
"Hongjie",
""
],
[
"Yuan",
"Yetian",
""
],
[
"Wang",
"Gang",
""
],
[
"Sun",
"Jian",
""
],
[
"Chen",
"Jie",
""
]
] | TITLE: Robust Offline Imitation Learning Through State-level Trajectory
Stitching
ABSTRACT: Imitation learning (IL) has proven effective for enabling robots to acquire
visuomotor skills through expert demonstrations. However, traditional IL
methods are limited by their reliance on high-quality, often scarce, expert
data, and suffer from covariate shift. To address these challenges, recent
advances in offline IL have incorporated suboptimal, unlabeled datasets into
the training. In this paper, we propose a novel approach to enhance policy
learning from mixed-quality offline datasets by leveraging task-relevant
trajectory fragments and rich environmental dynamics. Specifically, we
introduce a state-based search framework that stitches state-action pairs from
imperfect demonstrations, generating more diverse and informative training
trajectories. Experimental results on standard IL benchmarks and real-world
robotic tasks showcase that our proposed method significantly improves both
generalization and performance.
|
2503.22526 | Martin Ki\v{s}\v{s} | Martin Ki\v{s}\v{s} and Michal Hradi\v{s} and Martina
Dvo\v{r}\'akov\'a and V\'aclav Jirou\v{s}ek and Filip Kersch | AnnoPage Dataset: Dataset of Non-Textual Elements in Documents with
Fine-Grained Categorization | 15 pages, 2 tables, 6 figures; Submitted to ICDAR25 | null | null | null | cs.CV cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce the AnnoPage Dataset, a novel collection of 7550 pages from
historical documents, primarily in Czech and German, spanning from 1485 to the
present, focusing on the late 19th and early 20th centuries. The dataset is
designed to support research in document layout analysis and object detection.
Each page is annotated with axis-aligned bounding boxes (AABB) representing
elements of 25 categories of non-textual elements, such as images, maps,
decorative elements, or charts, following the Czech Methodology of image
document processing. The annotations were created by expert librarians to
ensure accuracy and consistency. The dataset also incorporates pages from
multiple, mainly historical, document datasets to enhance variability and
maintain continuity. The dataset is divided into development and test subsets,
with the test set carefully selected to maintain the category distribution. We
provide baseline results using YOLO and DETR object detectors, offering a
reference point for future research. The AnnoPage Dataset is publicly available
on Zenodo (https://doi.org/10.5281/zenodo.12788419), along with ground-truth
annotations in YOLO format.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 15:30:42 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Kišš",
"Martin",
""
],
[
"Hradiš",
"Michal",
""
],
[
"Dvořáková",
"Martina",
""
],
[
"Jiroušek",
"Václav",
""
],
[
"Kersch",
"Filip",
""
]
] | TITLE: AnnoPage Dataset: Dataset of Non-Textual Elements in Documents with
Fine-Grained Categorization
ABSTRACT: We introduce the AnnoPage Dataset, a novel collection of 7550 pages from
historical documents, primarily in Czech and German, spanning from 1485 to the
present, focusing on the late 19th and early 20th centuries. The dataset is
designed to support research in document layout analysis and object detection.
Each page is annotated with axis-aligned bounding boxes (AABB) representing
elements of 25 categories of non-textual elements, such as images, maps,
decorative elements, or charts, following the Czech Methodology of image
document processing. The annotations were created by expert librarians to
ensure accuracy and consistency. The dataset also incorporates pages from
multiple, mainly historical, document datasets to enhance variability and
maintain continuity. The dataset is divided into development and test subsets,
with the test set carefully selected to maintain the category distribution. We
provide baseline results using YOLO and DETR object detectors, offering a
reference point for future research. The AnnoPage Dataset is publicly available
on Zenodo (https://doi.org/10.5281/zenodo.12788419), along with ground-truth
annotations in YOLO format.
|
2503.22531 | Qisheng He | Qisheng He, Nicholas Summerfield, Peiyong Wang, Carri Glide-Hurst,
Ming Dong | Deterministic Medical Image Translation via High-fidelity Brownian
Bridges | null | null | null | null | eess.IV cs.CV cs.LG | http://creativecommons.org/licenses/by/4.0/ | Recent studies have shown that diffusion models produce superior synthetic
images when compared to Generative Adversarial Networks (GANs). However, their
outputs are often non-deterministic and lack high fidelity to the ground truth
due to the inherent randomness. In this paper, we propose a novel High-fidelity
Brownian bridge model (HiFi-BBrg) for deterministic medical image translations.
Our model comprises two distinct yet mutually beneficial mappings: a generation
mapping and a reconstruction mapping. The Brownian bridge training process is
guided by the fidelity loss and adversarial training in the reconstruction
mapping. This ensures that translated images can be accurately reversed to
their original forms, thereby achieving consistent translations with high
fidelity to the ground truth. Our extensive experiments on multiple datasets
show HiFi-BBrg outperforms state-of-the-art methods in multi-modal image
translation and multi-image super-resolution.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 15:33:28 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"He",
"Qisheng",
""
],
[
"Summerfield",
"Nicholas",
""
],
[
"Wang",
"Peiyong",
""
],
[
"Glide-Hurst",
"Carri",
""
],
[
"Dong",
"Ming",
""
]
] | TITLE: Deterministic Medical Image Translation via High-fidelity Brownian
Bridges
ABSTRACT: Recent studies have shown that diffusion models produce superior synthetic
images when compared to Generative Adversarial Networks (GANs). However, their
outputs are often non-deterministic and lack high fidelity to the ground truth
due to the inherent randomness. In this paper, we propose a novel High-fidelity
Brownian bridge model (HiFi-BBrg) for deterministic medical image translations.
Our model comprises two distinct yet mutually beneficial mappings: a generation
mapping and a reconstruction mapping. The Brownian bridge training process is
guided by the fidelity loss and adversarial training in the reconstruction
mapping. This ensures that translated images can be accurately reversed to
their original forms, thereby achieving consistent translations with high
fidelity to the ground truth. Our extensive experiments on multiple datasets
show HiFi-BBrg outperforms state-of-the-art methods in multi-modal image
translation and multi-image super-resolution.
|
2503.22537 | Remy Sabathier | Remy Sabathier, Niloy J. Mitra, David Novotny | LIM: Large Interpolator Model for Dynamic Reconstruction | null | null | null | null | cs.CV cs.AI | http://creativecommons.org/licenses/by/4.0/ | Reconstructing dynamic assets from video data is central to many in computer
vision and graphics tasks. Existing 4D reconstruction approaches are limited by
category-specific models or slow optimization-based methods. Inspired by the
recent Large Reconstruction Model (LRM), we present the Large Interpolation
Model (LIM), a transformer-based feed-forward solution, guided by a novel
causal consistency loss, for interpolating implicit 3D representations across
time. Given implicit 3D representations at times $t_0$ and $t_1$, LIM produces
a deformed shape at any continuous time $t\in[t_0,t_1]$, delivering
high-quality interpolated frames in seconds. Furthermore, LIM allows explicit
mesh tracking across time, producing a consistently uv-textured mesh sequence
ready for integration into existing production pipelines. We also use LIM, in
conjunction with a diffusion-based multiview generator, to produce dynamic 4D
reconstructions from monocular videos. We evaluate LIM on various dynamic
datasets, benchmarking against image-space interpolation methods (e.g., FiLM)
and direct triplane linear interpolation, and demonstrate clear advantages. In
summary, LIM is the first feed-forward model capable of high-speed tracked 4D
asset reconstruction across diverse categories.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 15:36:53 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Sabathier",
"Remy",
""
],
[
"Mitra",
"Niloy J.",
""
],
[
"Novotny",
"David",
""
]
] | TITLE: LIM: Large Interpolator Model for Dynamic Reconstruction
ABSTRACT: Reconstructing dynamic assets from video data is central to many in computer
vision and graphics tasks. Existing 4D reconstruction approaches are limited by
category-specific models or slow optimization-based methods. Inspired by the
recent Large Reconstruction Model (LRM), we present the Large Interpolation
Model (LIM), a transformer-based feed-forward solution, guided by a novel
causal consistency loss, for interpolating implicit 3D representations across
time. Given implicit 3D representations at times $t_0$ and $t_1$, LIM produces
a deformed shape at any continuous time $t\in[t_0,t_1]$, delivering
high-quality interpolated frames in seconds. Furthermore, LIM allows explicit
mesh tracking across time, producing a consistently uv-textured mesh sequence
ready for integration into existing production pipelines. We also use LIM, in
conjunction with a diffusion-based multiview generator, to produce dynamic 4D
reconstructions from monocular videos. We evaluate LIM on various dynamic
datasets, benchmarking against image-space interpolation methods (e.g., FiLM)
and direct triplane linear interpolation, and demonstrate clear advantages. In
summary, LIM is the first feed-forward model capable of high-speed tracked 4D
asset reconstruction across diverse categories.
|
2503.22539 | Yijun Quan | Yijun Quan, Zushu Li, Giovanni Montana | Efficient Verified Machine Unlearning For Distillation | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Growing data privacy demands, driven by regulations like GDPR and CCPA,
require machine unlearning methods capable of swiftly removing the influence of
specific training points. Although verified approaches like SISA, using data
slicing and checkpointing, achieve efficient unlearning for single models by
reverting to intermediate states, these methods struggle in teacher-student
knowledge distillation settings. Unlearning in the teacher typically forces
costly, complete student retraining due to pervasive information propagation
during distillation. Our primary contribution is PURGE (Partitioned Unlearning
with Retraining Guarantee for Ensembles), a novel framework integrating
verified unlearning with distillation. We introduce constituent mapping and an
incremental multi-teacher strategy that partitions the distillation process,
confines each teacher constituent's impact to distinct student data subsets,
and crucially maintains data isolation. The PURGE framework substantially
reduces retraining overhead, requiring only partial student updates when
teacher-side unlearning occurs. We provide both theoretical analysis,
quantifying significant speed-ups in the unlearning process, and empirical
validation on multiple datasets, demonstrating that PURGE achieves these
efficiency gains while maintaining student accuracy comparable to standard
baselines.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 15:38:07 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Quan",
"Yijun",
""
],
[
"Li",
"Zushu",
""
],
[
"Montana",
"Giovanni",
""
]
] | TITLE: Efficient Verified Machine Unlearning For Distillation
ABSTRACT: Growing data privacy demands, driven by regulations like GDPR and CCPA,
require machine unlearning methods capable of swiftly removing the influence of
specific training points. Although verified approaches like SISA, using data
slicing and checkpointing, achieve efficient unlearning for single models by
reverting to intermediate states, these methods struggle in teacher-student
knowledge distillation settings. Unlearning in the teacher typically forces
costly, complete student retraining due to pervasive information propagation
during distillation. Our primary contribution is PURGE (Partitioned Unlearning
with Retraining Guarantee for Ensembles), a novel framework integrating
verified unlearning with distillation. We introduce constituent mapping and an
incremental multi-teacher strategy that partitions the distillation process,
confines each teacher constituent's impact to distinct student data subsets,
and crucially maintains data isolation. The PURGE framework substantially
reduces retraining overhead, requiring only partial student updates when
teacher-side unlearning occurs. We provide both theoretical analysis,
quantifying significant speed-ups in the unlearning process, and empirical
validation on multiple datasets, demonstrating that PURGE achieves these
efficiency gains while maintaining student accuracy comparable to standard
baselines.
|
2503.22541 | Haicheng Liao | Haicheng Liao, Hanlin Kong, Bin Rao, Bonan Wang, Chengyue Wang, Guyang
Yu, Yuming Huang, Ruru Tang, Chengzhong Xu, and Zhenning Li | SafeCast: Risk-Responsive Motion Forecasting for Autonomous Vehicles | null | null | null | null | cs.RO cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Accurate motion forecasting is essential for the safety and reliability of
autonomous driving (AD) systems. While existing methods have made significant
progress, they often overlook explicit safety constraints and struggle to
capture the complex interactions among traffic agents, environmental factors,
and motion dynamics. To address these challenges, we present SafeCast, a
risk-responsive motion forecasting model that integrates safety-aware
decision-making with uncertainty-aware adaptability. SafeCast is the first to
incorporate the Responsibility-Sensitive Safety (RSS) framework into motion
forecasting, encoding interpretable safety rules--such as safe distances and
collision avoidance--based on traffic norms and physical principles. To further
enhance robustness, we introduce the Graph Uncertainty Feature (GUF), a
graph-based module that injects learnable noise into Graph Attention Networks,
capturing real-world uncertainties and enhancing generalization across diverse
scenarios. We evaluate SafeCast on four real-world benchmark datasets--Next
Generation Simulation (NGSIM), Highway Drone (HighD), ApolloScape, and the
Macao Connected Autonomous Driving (MoCAD)--covering highway, urban, and
mixed-autonomy traffic environments. Our model achieves state-of-the-art (SOTA)
accuracy while maintaining a lightweight architecture and low inference
latency, underscoring its potential for real-time deployment in safety-critical
AD systems.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 15:38:21 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Liao",
"Haicheng",
""
],
[
"Kong",
"Hanlin",
""
],
[
"Rao",
"Bin",
""
],
[
"Wang",
"Bonan",
""
],
[
"Wang",
"Chengyue",
""
],
[
"Yu",
"Guyang",
""
],
[
"Huang",
"Yuming",
""
],
[
"Tang",
"Ruru",
""
],
[
"Xu",
"Chengzhong",
""
],
[
"Li",
"Zhenning",
""
]
] | TITLE: SafeCast: Risk-Responsive Motion Forecasting for Autonomous Vehicles
ABSTRACT: Accurate motion forecasting is essential for the safety and reliability of
autonomous driving (AD) systems. While existing methods have made significant
progress, they often overlook explicit safety constraints and struggle to
capture the complex interactions among traffic agents, environmental factors,
and motion dynamics. To address these challenges, we present SafeCast, a
risk-responsive motion forecasting model that integrates safety-aware
decision-making with uncertainty-aware adaptability. SafeCast is the first to
incorporate the Responsibility-Sensitive Safety (RSS) framework into motion
forecasting, encoding interpretable safety rules--such as safe distances and
collision avoidance--based on traffic norms and physical principles. To further
enhance robustness, we introduce the Graph Uncertainty Feature (GUF), a
graph-based module that injects learnable noise into Graph Attention Networks,
capturing real-world uncertainties and enhancing generalization across diverse
scenarios. We evaluate SafeCast on four real-world benchmark datasets--Next
Generation Simulation (NGSIM), Highway Drone (HighD), ApolloScape, and the
Macao Connected Autonomous Driving (MoCAD)--covering highway, urban, and
mixed-autonomy traffic environments. Our model achieves state-of-the-art (SOTA)
accuracy while maintaining a lightweight architecture and low inference
latency, underscoring its potential for real-time deployment in safety-critical
AD systems.
|
2503.22557 | Zhendi Gong | Zhendi Gong, Susan Francis, Eleanor Cox, Stamatios N. Sotiropoulos,
Dorothee P. Auer, Guoping Qiu, Andrew P. French, Xin Chen | MO-CTranS: A unified multi-organ segmentation model learning from
multiple heterogeneously labelled datasets | Accepted by International Symposium on Biomedical Imaging (ISIB) 2025
as an oral presentation | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Multi-organ segmentation holds paramount significance in many clinical tasks.
In practice, compared to large fully annotated datasets, multiple small
datasets are often more accessible and organs are not labelled consistently.
Normally, an individual model is trained for each of these datasets, which is
not an effective way of using data for model learning. It remains challenging
to train a single model that can robustly learn from several partially labelled
datasets due to label conflict and data imbalance problems. We propose
MO-CTranS: a single model that can overcome such problems. MO-CTranS contains a
CNN-based encoder and a Transformer-based decoder, which are connected in a
multi-resolution manner. Task-specific tokens are introduced in the decoder to
help differentiate label discrepancies. Our method was evaluated and compared
to several baseline models and state-of-the-art (SOTA) solutions on abdominal
MRI datasets that were acquired in different views (i.e. axial and coronal) and
annotated for different organs (i.e. liver, kidney, spleen). Our method
achieved better performance (most were statistically significant) than the
compared methods. Github link: https://github.com/naisops/MO-CTranS.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 16:00:59 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Gong",
"Zhendi",
""
],
[
"Francis",
"Susan",
""
],
[
"Cox",
"Eleanor",
""
],
[
"Sotiropoulos",
"Stamatios N.",
""
],
[
"Auer",
"Dorothee P.",
""
],
[
"Qiu",
"Guoping",
""
],
[
"French",
"Andrew P.",
""
],
[
"Chen",
"Xin",
""
]
] | TITLE: MO-CTranS: A unified multi-organ segmentation model learning from
multiple heterogeneously labelled datasets
ABSTRACT: Multi-organ segmentation holds paramount significance in many clinical tasks.
In practice, compared to large fully annotated datasets, multiple small
datasets are often more accessible and organs are not labelled consistently.
Normally, an individual model is trained for each of these datasets, which is
not an effective way of using data for model learning. It remains challenging
to train a single model that can robustly learn from several partially labelled
datasets due to label conflict and data imbalance problems. We propose
MO-CTranS: a single model that can overcome such problems. MO-CTranS contains a
CNN-based encoder and a Transformer-based decoder, which are connected in a
multi-resolution manner. Task-specific tokens are introduced in the decoder to
help differentiate label discrepancies. Our method was evaluated and compared
to several baseline models and state-of-the-art (SOTA) solutions on abdominal
MRI datasets that were acquired in different views (i.e. axial and coronal) and
annotated for different organs (i.e. liver, kidney, spleen). Our method
achieved better performance (most were statistically significant) than the
compared methods. Github link: https://github.com/naisops/MO-CTranS.
|
2503.22563 | Andrea Sebastiani | Pasquale Cascarano, Lorenzo Stacchio, Andrea Sebastiani, Alessandro
Benfenati, Ulugbek S. Kamilov, Gustavo Marfia | RELD: Regularization by Latent Diffusion Models for Image Restoration | null | null | null | null | eess.IV cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In recent years, Diffusion Models have become the new state-of-the-art in
deep generative modeling, ending the long-time dominance of Generative
Adversarial Networks. Inspired by the Regularization by Denoising principle, we
introduce an approach that integrates a Latent Diffusion Model, trained for the
denoising task, into a variational framework using Half-Quadratic Splitting,
exploiting its regularization properties. This approach, under appropriate
conditions that can be easily met in various imaging applications, allows for
reduced computational cost while achieving high-quality results. The proposed
strategy, called Regularization by Latent Denoising (RELD), is then tested on a
dataset of natural images, for image denoising, deblurring, and
super-resolution tasks. The numerical experiments show that RELD is competitive
with other state-of-the-art methods, particularly achieving remarkable results
when evaluated using perceptual quality metrics.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 16:04:21 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Cascarano",
"Pasquale",
""
],
[
"Stacchio",
"Lorenzo",
""
],
[
"Sebastiani",
"Andrea",
""
],
[
"Benfenati",
"Alessandro",
""
],
[
"Kamilov",
"Ulugbek S.",
""
],
[
"Marfia",
"Gustavo",
""
]
] | TITLE: RELD: Regularization by Latent Diffusion Models for Image Restoration
ABSTRACT: In recent years, Diffusion Models have become the new state-of-the-art in
deep generative modeling, ending the long-time dominance of Generative
Adversarial Networks. Inspired by the Regularization by Denoising principle, we
introduce an approach that integrates a Latent Diffusion Model, trained for the
denoising task, into a variational framework using Half-Quadratic Splitting,
exploiting its regularization properties. This approach, under appropriate
conditions that can be easily met in various imaging applications, allows for
reduced computational cost while achieving high-quality results. The proposed
strategy, called Regularization by Latent Denoising (RELD), is then tested on a
dataset of natural images, for image denoising, deblurring, and
super-resolution tasks. The numerical experiments show that RELD is competitive
with other state-of-the-art methods, particularly achieving remarkable results
when evaluated using perceptual quality metrics.
|
2503.22569 | Barbara Alexandra Hoffmann | Barbara Hoffmann, Ruben Mayer | Comparing Methods for Bias Mitigation in Graph Neural Networks | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | This paper examines the critical role of Graph Neural Networks (GNNs) in data
preparation for generative artificial intelligence (GenAI) systems, with a
particular focus on addressing and mitigating biases. We present a comparative
analysis of three distinct methods for bias mitigation: data sparsification,
feature modification, and synthetic data augmentation. Through experimental
analysis using the german credit dataset, we evaluate these approaches using
multiple fairness metrics, including statistical parity, equality of
opportunity, and false positive rates. Our research demonstrates that while all
methods improve fairness metrics compared to the original dataset, stratified
sampling and synthetic data augmentation using GraphSAGE prove particularly
effective in balancing demographic representation while maintaining model
performance. The results provide practical insights for developing more
equitable AI systems while maintaining model performance.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 16:18:48 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Hoffmann",
"Barbara",
""
],
[
"Mayer",
"Ruben",
""
]
] | TITLE: Comparing Methods for Bias Mitigation in Graph Neural Networks
ABSTRACT: This paper examines the critical role of Graph Neural Networks (GNNs) in data
preparation for generative artificial intelligence (GenAI) systems, with a
particular focus on addressing and mitigating biases. We present a comparative
analysis of three distinct methods for bias mitigation: data sparsification,
feature modification, and synthetic data augmentation. Through experimental
analysis using the german credit dataset, we evaluate these approaches using
multiple fairness metrics, including statistical parity, equality of
opportunity, and false positive rates. Our research demonstrates that while all
methods improve fairness metrics compared to the original dataset, stratified
sampling and synthetic data augmentation using GraphSAGE prove particularly
effective in balancing demographic representation while maintaining model
performance. The results provide practical insights for developing more
equitable AI systems while maintaining model performance.
|
2503.22582 | Sarubi Thillainathan | Sarubi Thillainathan, Songchen Yuan, En-Shiun Annie Lee, Sanath
Jayasena, Surangika Ranathunga | Beyond Vanilla Fine-Tuning: Leveraging Multistage, Multilingual, and
Domain-Specific Methods for Low-Resource Machine Translation | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Fine-tuning multilingual sequence-to-sequence large language models (msLLMs)
has shown promise in developing neural machine translation (NMT) systems for
low-resource languages (LRLs). However, conventional single-stage fine-tuning
methods struggle in extremely low-resource NMT settings, where training data is
very limited. This paper contributes to artificial intelligence by proposing
two approaches for adapting msLLMs in these challenging scenarios: (1)
continual pre-training (CPT), where the msLLM is further trained with
domain-specific monolingual data to compensate for the under-representation of
LRLs, and (2) intermediate task transfer learning (ITTL), a method that
fine-tunes the msLLM with both in-domain and out-of-domain parallel data to
enhance its translation capabilities across various domains and tasks. As an
application in engineering, these methods are implemented in NMT systems for
Sinhala, Tamil, and English (six language pairs) in domain-specific, extremely
low-resource settings (datasets containing fewer than 100,000 samples). Our
experiments reveal that these approaches enhance translation performance by an
average of +1.47 bilingual evaluation understudy (BLEU) score compared to the
standard single-stage fine-tuning baseline across all translation directions.
Additionally, a multi-model ensemble further improves performance by an
additional BLEU score.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 16:30:28 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Thillainathan",
"Sarubi",
""
],
[
"Yuan",
"Songchen",
""
],
[
"Lee",
"En-Shiun Annie",
""
],
[
"Jayasena",
"Sanath",
""
],
[
"Ranathunga",
"Surangika",
""
]
] | TITLE: Beyond Vanilla Fine-Tuning: Leveraging Multistage, Multilingual, and
Domain-Specific Methods for Low-Resource Machine Translation
ABSTRACT: Fine-tuning multilingual sequence-to-sequence large language models (msLLMs)
has shown promise in developing neural machine translation (NMT) systems for
low-resource languages (LRLs). However, conventional single-stage fine-tuning
methods struggle in extremely low-resource NMT settings, where training data is
very limited. This paper contributes to artificial intelligence by proposing
two approaches for adapting msLLMs in these challenging scenarios: (1)
continual pre-training (CPT), where the msLLM is further trained with
domain-specific monolingual data to compensate for the under-representation of
LRLs, and (2) intermediate task transfer learning (ITTL), a method that
fine-tunes the msLLM with both in-domain and out-of-domain parallel data to
enhance its translation capabilities across various domains and tasks. As an
application in engineering, these methods are implemented in NMT systems for
Sinhala, Tamil, and English (six language pairs) in domain-specific, extremely
low-resource settings (datasets containing fewer than 100,000 samples). Our
experiments reveal that these approaches enhance translation performance by an
average of +1.47 bilingual evaluation understudy (BLEU) score compared to the
standard single-stage fine-tuning baseline across all translation directions.
Additionally, a multi-model ensemble further improves performance by an
additional BLEU score.
|
2503.22585 | Laura Manrique-G\'omez | Kevin Cohen, Laura Manrique-G\'omez, Rub\'en Manrique | Historical Ink: Exploring Large Language Models for Irony Detection in
19th-Century Spanish | null | null | null | null | cs.CL cs.AI cs.DL | http://creativecommons.org/licenses/by/4.0/ | This study explores the use of large language models (LLMs) to enhance
datasets and improve irony detection in 19th-century Latin American newspapers.
Two strategies were employed to evaluate the efficacy of BERT and GPT-4o models
in capturing the subtle nuances nature of irony, through both multi-class and
binary classification tasks. First, we implemented dataset enhancements focused
on enriching emotional and contextual cues; however, these showed limited
impact on historical language analysis. The second strategy, a semi-automated
annotation process, effectively addressed class imbalance and augmented the
dataset with high-quality annotations. Despite the challenges posed by the
complexity of irony, this work contributes to the advancement of sentiment
analysis through two key contributions: introducing a new historical Spanish
dataset tagged for sentiment analysis and irony detection, and proposing a
semi-automated annotation methodology where human expertise is crucial for
refining LLMs results, enriched by incorporating historical and cultural
contexts as core features.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 16:33:24 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Cohen",
"Kevin",
""
],
[
"Manrique-Gómez",
"Laura",
""
],
[
"Manrique",
"Rubén",
""
]
] | TITLE: Historical Ink: Exploring Large Language Models for Irony Detection in
19th-Century Spanish
ABSTRACT: This study explores the use of large language models (LLMs) to enhance
datasets and improve irony detection in 19th-century Latin American newspapers.
Two strategies were employed to evaluate the efficacy of BERT and GPT-4o models
in capturing the subtle nuances nature of irony, through both multi-class and
binary classification tasks. First, we implemented dataset enhancements focused
on enriching emotional and contextual cues; however, these showed limited
impact on historical language analysis. The second strategy, a semi-automated
annotation process, effectively addressed class imbalance and augmented the
dataset with high-quality annotations. Despite the challenges posed by the
complexity of irony, this work contributes to the advancement of sentiment
analysis through two key contributions: introducing a new historical Spanish
dataset tagged for sentiment analysis and irony detection, and proposing a
semi-automated annotation methodology where human expertise is crucial for
refining LLMs results, enriched by incorporating historical and cultural
contexts as core features.
|
2503.22589 | Bryce Dietrich | Adam Breuer, Bryce J. Dietrich, Michael H. Crespin, Matthew Butler,
J.A. Pyrse, and Kosuke Imai | Using AI to Summarize US Presidential Campaign TV Advertisement Videos,
1952-2012 | 17 pages, 7 tables, 4 figures, and linked datasets | null | null | null | cs.MM cs.AI cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper introduces the largest and most comprehensive dataset of US
presidential campaign television advertisements, available in digital format.
The dataset also includes machine-searchable transcripts and high-quality
summaries designed to facilitate a variety of academic research. To date, there
has been great interest in collecting and analyzing US presidential campaign
advertisements, but the need for manual procurement and annotation led many to
rely on smaller subsets. We design a large-scale parallelized, AI-based
analysis pipeline that automates the laborious process of preparing,
transcribing, and summarizing videos. We then apply this methodology to the
9,707 presidential ads from the Julian P. Kanter Political Commercial Archive.
We conduct extensive human evaluations to show that these transcripts and
summaries match the quality of manually generated alternatives. We illustrate
the value of this data by including an application that tracks the genesis and
evolution of current focal issue areas over seven decades of presidential
elections. Our analysis pipeline and codebase also show how to use LLM-based
tools to obtain high-quality summaries for other video datasets.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 16:36:23 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Breuer",
"Adam",
""
],
[
"Dietrich",
"Bryce J.",
""
],
[
"Crespin",
"Michael H.",
""
],
[
"Butler",
"Matthew",
""
],
[
"Pyrse",
"J. A.",
""
],
[
"Imai",
"Kosuke",
""
]
] | TITLE: Using AI to Summarize US Presidential Campaign TV Advertisement Videos,
1952-2012
ABSTRACT: This paper introduces the largest and most comprehensive dataset of US
presidential campaign television advertisements, available in digital format.
The dataset also includes machine-searchable transcripts and high-quality
summaries designed to facilitate a variety of academic research. To date, there
has been great interest in collecting and analyzing US presidential campaign
advertisements, but the need for manual procurement and annotation led many to
rely on smaller subsets. We design a large-scale parallelized, AI-based
analysis pipeline that automates the laborious process of preparing,
transcribing, and summarizing videos. We then apply this methodology to the
9,707 presidential ads from the Julian P. Kanter Political Commercial Archive.
We conduct extensive human evaluations to show that these transcripts and
summaries match the quality of manually generated alternatives. We illustrate
the value of this data by including an application that tracks the genesis and
evolution of current focal issue areas over seven decades of presidential
elections. Our analysis pipeline and codebase also show how to use LLM-based
tools to obtain high-quality summaries for other video datasets.
|
2503.22592 | Thomas Boucher | Thomas Boucher, Nicholas Tetlow, Annie Fung, Amy Dewar, Pietro Arina,
Sven Kerneis, John Whittle, Evangelos B. Mazomenos | KEVS: Enhancing Segmentation of Visceral Adipose Tissue in
Pre-Cystectomy CT with Gaussian Kernel Density Estimation | Preprint for submission to IPCAI special edition of IJCARS 2025,
version prior to any peer review | null | null | null | eess.IV cs.AI cs.CV | http://creativecommons.org/licenses/by/4.0/ | Purpose: The distribution of visceral adipose tissue (VAT) in cystectomy
patients is indicative of the incidence of post-operative complications.
Existing VAT segmentation methods for computed tomography (CT) employing
intensity thresholding have limitations relating to inter-observer variability.
Moreover, the difficulty in creating ground-truth masks limits the development
of deep learning (DL) models for this task. This paper introduces a novel
method for VAT prediction in pre-cystectomy CT, which is fully automated and
does not require ground-truth VAT masks for training, overcoming aforementioned
limitations. Methods: We introduce the Kernel density Enhanced VAT Segmentator
( KEVS), combining a DL semantic segmentation model, for multi-body feature
prediction, with Gaussian kernel density estimation analysis of predicted
subcutaneous adipose tissue to achieve accurate scan-specific predictions of
VAT in the abdominal cavity. Uniquely for a DL pipeline, KEVS does not require
ground-truth VAT masks. Results: We verify the ability of KEVS to accurately
segment abdominal organs in unseen CT data and compare KEVS VAT segmentation
predictions to existing state-of-the-art (SOTA) approaches in a dataset of 20
pre-cystectomy CT scans, collected from University College London Hospital
(UCLH-Cyst), with expert ground-truth annotations. KEVS presents a 4.80% and
6.02% improvement in Dice Coefficient over the second best DL and
thresholding-based VAT segmentation techniques respectively when evaluated on
UCLH-Cyst. Conclusion: This research introduces KEVS; an automated, SOTA method
for the prediction of VAT in pre-cystectomy CT which eliminates inter-observer
variability and is trained entirely on open-source CT datasets which do not
contain ground-truth VAT masks.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 16:41:09 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Boucher",
"Thomas",
""
],
[
"Tetlow",
"Nicholas",
""
],
[
"Fung",
"Annie",
""
],
[
"Dewar",
"Amy",
""
],
[
"Arina",
"Pietro",
""
],
[
"Kerneis",
"Sven",
""
],
[
"Whittle",
"John",
""
],
[
"Mazomenos",
"Evangelos B.",
""
]
] | TITLE: KEVS: Enhancing Segmentation of Visceral Adipose Tissue in
Pre-Cystectomy CT with Gaussian Kernel Density Estimation
ABSTRACT: Purpose: The distribution of visceral adipose tissue (VAT) in cystectomy
patients is indicative of the incidence of post-operative complications.
Existing VAT segmentation methods for computed tomography (CT) employing
intensity thresholding have limitations relating to inter-observer variability.
Moreover, the difficulty in creating ground-truth masks limits the development
of deep learning (DL) models for this task. This paper introduces a novel
method for VAT prediction in pre-cystectomy CT, which is fully automated and
does not require ground-truth VAT masks for training, overcoming aforementioned
limitations. Methods: We introduce the Kernel density Enhanced VAT Segmentator
( KEVS), combining a DL semantic segmentation model, for multi-body feature
prediction, with Gaussian kernel density estimation analysis of predicted
subcutaneous adipose tissue to achieve accurate scan-specific predictions of
VAT in the abdominal cavity. Uniquely for a DL pipeline, KEVS does not require
ground-truth VAT masks. Results: We verify the ability of KEVS to accurately
segment abdominal organs in unseen CT data and compare KEVS VAT segmentation
predictions to existing state-of-the-art (SOTA) approaches in a dataset of 20
pre-cystectomy CT scans, collected from University College London Hospital
(UCLH-Cyst), with expert ground-truth annotations. KEVS presents a 4.80% and
6.02% improvement in Dice Coefficient over the second best DL and
thresholding-based VAT segmentation techniques respectively when evaluated on
UCLH-Cyst. Conclusion: This research introduces KEVS; an automated, SOTA method
for the prediction of VAT in pre-cystectomy CT which eliminates inter-observer
variability and is trained entirely on open-source CT datasets which do not
contain ground-truth VAT masks.
|
2503.22594 | Philipp Schaer | Dirk Tunger and Philipp Schaer | On the Alignment of Post-Publication Reviews & Bibliometric and
Altmetric Impact -- A Case Study on Expert Statements from the Science Media
Center Germany | Accepted at The First Workshop on Scholarly Information Access
(SCOLIA) | null | null | null | cs.DL | http://creativecommons.org/licenses/by-sa/4.0/ | In the context of academic publishing and peer review, this study
investigates the relationship between post-publication expert evaluations,
their agreement levels, and the subsequent scientific and public recognition of
the reviewed research. Using expert statements from the Science Media Center
Germany as a dataset, we analyze Research in Context reviews to examine the
alignment between qualitative post-publication assessments and bibliometric as
well as altmetric indicators. We employ a Large Language Model to translate
unstructured expert reviews into a structured rating scheme. Furthermore, we
correlate these evaluations with citation counts from the Web of Science and
alternative impact metrics such as the Altmetric Attention Score, news
mentions, and Mendeley readership statistics from the Altmetric Explorer. We
investigate the alignment of positive or critical post-publication reviews and
high or low citation or altmetric counts.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 16:41:41 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Tunger",
"Dirk",
""
],
[
"Schaer",
"Philipp",
""
]
] | TITLE: On the Alignment of Post-Publication Reviews & Bibliometric and
Altmetric Impact -- A Case Study on Expert Statements from the Science Media
Center Germany
ABSTRACT: In the context of academic publishing and peer review, this study
investigates the relationship between post-publication expert evaluations,
their agreement levels, and the subsequent scientific and public recognition of
the reviewed research. Using expert statements from the Science Media Center
Germany as a dataset, we analyze Research in Context reviews to examine the
alignment between qualitative post-publication assessments and bibliometric as
well as altmetric indicators. We employ a Large Language Model to translate
unstructured expert reviews into a structured rating scheme. Furthermore, we
correlate these evaluations with citation counts from the Web of Science and
alternative impact metrics such as the Altmetric Attention Score, news
mentions, and Mendeley readership statistics from the Altmetric Explorer. We
investigate the alignment of positive or critical post-publication reviews and
high or low citation or altmetric counts.
|
2503.22595 | Steven McClendon | S. Aaron McClendon, Vishaal Venkatesh, Juan Morinelli | Reinforcement Learning for Machine Learning Model Deployment: Evaluating
Multi-Armed Bandits in ML Ops Environments | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | In modern ML Ops environments, model deployment is a critical process that
traditionally relies on static heuristics such as validation error comparisons
and A/B testing. However, these methods require human intervention to adapt to
real-world deployment challenges, such as model drift or unexpected performance
degradation. We investigate whether reinforcement learning, specifically
multi-armed bandit (MAB) algorithms, can dynamically manage model deployment
decisions more effectively. Our approach enables more adaptive production
environments by continuously evaluating deployed models and rolling back
underperforming ones in real-time. We test six model selection strategies
across two real-world datasets and find that RL based approaches match or
exceed traditional methods in performance. Our findings suggest that
reinforcement learning (RL)-based model management can improve automation,
reduce reliance on manual interventions, and mitigate risks associated with
post-deployment model failures.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 16:42:21 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"McClendon",
"S. Aaron",
""
],
[
"Venkatesh",
"Vishaal",
""
],
[
"Morinelli",
"Juan",
""
]
] | TITLE: Reinforcement Learning for Machine Learning Model Deployment: Evaluating
Multi-Armed Bandits in ML Ops Environments
ABSTRACT: In modern ML Ops environments, model deployment is a critical process that
traditionally relies on static heuristics such as validation error comparisons
and A/B testing. However, these methods require human intervention to adapt to
real-world deployment challenges, such as model drift or unexpected performance
degradation. We investigate whether reinforcement learning, specifically
multi-armed bandit (MAB) algorithms, can dynamically manage model deployment
decisions more effectively. Our approach enables more adaptive production
environments by continuously evaluating deployed models and rolling back
underperforming ones in real-time. We test six model selection strategies
across two real-world datasets and find that RL based approaches match or
exceed traditional methods in performance. Our findings suggest that
reinforcement learning (RL)-based model management can improve automation,
reduce reliance on manual interventions, and mitigate risks associated with
post-deployment model failures.
|
2503.22629 | Stefano Grassi | Stefano Grassi | Sentiment Classification of Thai Central Bank Press Releases Using
Supervised Learning | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Central bank communication plays a critical role in shaping economic
expectations and monetary policy effectiveness. This study applies supervised
machine learning techniques to classify the sentiment of press releases from
the Bank of Thailand, addressing gaps in research that primarily focus on
lexicon-based approaches. My findings show that supervised learning can be an
effective method, even with smaller datasets, and serves as a starting point
for further automation. However, achieving higher accuracy and better
generalization requires a substantial amount of labeled data, which is
time-consuming and demands expertise. Using models such as Na\"ive Bayes,
Random Forest and SVM, this study demonstrates the applicability of machine
learning for central bank sentiment analysis, with English-language
communications from the Thai Central Bank as a case study.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 17:20:41 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Grassi",
"Stefano",
""
]
] | TITLE: Sentiment Classification of Thai Central Bank Press Releases Using
Supervised Learning
ABSTRACT: Central bank communication plays a critical role in shaping economic
expectations and monetary policy effectiveness. This study applies supervised
machine learning techniques to classify the sentiment of press releases from
the Bank of Thailand, addressing gaps in research that primarily focus on
lexicon-based approaches. My findings show that supervised learning can be an
effective method, even with smaller datasets, and serves as a starting point
for further automation. However, achieving higher accuracy and better
generalization requires a substantial amount of labeled data, which is
time-consuming and demands expertise. Using models such as Na\"ive Bayes,
Random Forest and SVM, this study demonstrates the applicability of machine
learning for central bank sentiment analysis, with English-language
communications from the Thai Central Bank as a case study.
|
2503.22634 | Adam Wei | Adam Wei, Abhinav Agarwal, Boyuan Chen, Rohan Bosworth, Nicholas
Pfaff, Russ Tedrake | Empirical Analysis of Sim-and-Real Cotraining Of Diffusion Policies For
Planar Pushing from Pixels | 9 pages, 15 figures, In Submission to IROS 2025 | null | null | null | cs.RO cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In imitation learning for robotics, cotraining with demonstration data
generated both in simulation and on real hardware has emerged as a powerful
recipe to overcome the sim2real gap. This work seeks to elucidate basic
principles of this sim-and-real cotraining to help inform simulation design,
sim-and-real dataset creation, and policy training. Focusing narrowly on the
canonical task of planar pushing from camera inputs enabled us to be thorough
in our study. These experiments confirm that cotraining with simulated data
\emph{can} dramatically improve performance in real, especially when real data
is limited. Performance gains scale with simulated data, but eventually
plateau; real-world data increases this performance ceiling. The results also
suggest that reducing the domain gap in physics may be more important than
visual fidelity for non-prehensile manipulation tasks. Perhaps surprisingly,
having some visual domain gap actually helps the cotrained policy -- binary
probes reveal that high-performing policies learn to distinguish simulated
domains from real. We conclude by investigating this nuance and mechanisms that
facilitate positive transfer between sim-and-real. In total, our experiments
span over 40 real-world policies (evaluated on 800+ trials) and 200 simulated
policies (evaluated on 40,000+ trials).
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 17:25:57 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Wei",
"Adam",
""
],
[
"Agarwal",
"Abhinav",
""
],
[
"Chen",
"Boyuan",
""
],
[
"Bosworth",
"Rohan",
""
],
[
"Pfaff",
"Nicholas",
""
],
[
"Tedrake",
"Russ",
""
]
] | TITLE: Empirical Analysis of Sim-and-Real Cotraining Of Diffusion Policies For
Planar Pushing from Pixels
ABSTRACT: In imitation learning for robotics, cotraining with demonstration data
generated both in simulation and on real hardware has emerged as a powerful
recipe to overcome the sim2real gap. This work seeks to elucidate basic
principles of this sim-and-real cotraining to help inform simulation design,
sim-and-real dataset creation, and policy training. Focusing narrowly on the
canonical task of planar pushing from camera inputs enabled us to be thorough
in our study. These experiments confirm that cotraining with simulated data
\emph{can} dramatically improve performance in real, especially when real data
is limited. Performance gains scale with simulated data, but eventually
plateau; real-world data increases this performance ceiling. The results also
suggest that reducing the domain gap in physics may be more important than
visual fidelity for non-prehensile manipulation tasks. Perhaps surprisingly,
having some visual domain gap actually helps the cotrained policy -- binary
probes reveal that high-performing policies learn to distinguish simulated
domains from real. We conclude by investigating this nuance and mechanisms that
facilitate positive transfer between sim-and-real. In total, our experiments
span over 40 real-world policies (evaluated on 800+ trials) and 200 simulated
policies (evaluated on 40,000+ trials).
|
2503.22655 | Xiaomin Yu | Xiaomin Yu, Pengxiang Ding, Wenjie Zhang, Siteng Huang, Songyang Gao,
Chengwei Qin, Kejian Wu, Zhaoxin Fan, Ziyue Qiao, Donglin Wang | Unicorn: Text-Only Data Synthesis for Vision Language Model Training | null | null | null | null | cs.AI cs.CV cs.MM | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Training vision-language models (VLMs) typically requires large-scale,
high-quality image-text pairs, but collecting or synthesizing such data is
costly. In contrast, text data is abundant and inexpensive, prompting the
question: can high-quality multimodal training data be synthesized purely from
text? To tackle this, we propose a cross-integrated three-stage multimodal data
synthesis framework, which generates two datasets: Unicorn-1.2M and
Unicorn-471K-Instruction. In Stage 1: Diverse Caption Data Synthesis, we
construct 1.2M semantically diverse high-quality captions by expanding sparse
caption seeds using large language models (LLMs). In Stage 2:
Instruction-Tuning Data Generation, we further process 471K captions into
multi-turn instruction-tuning tasks to support complex reasoning. Finally, in
Stage 3: Modality Representation Transfer, these textual captions
representations are transformed into visual representations, resulting in
diverse synthetic image representations. This three-stage process enables us to
construct Unicorn-1.2M for pretraining and Unicorn-471K-Instruction for
instruction-tuning, without relying on real images. By eliminating the
dependency on real images while maintaining data quality and diversity, our
framework offers a cost-effective and scalable solution for VLMs training. Code
is available at https://github.com/Yu-xm/Unicorn.git.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 17:43:00 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Yu",
"Xiaomin",
""
],
[
"Ding",
"Pengxiang",
""
],
[
"Zhang",
"Wenjie",
""
],
[
"Huang",
"Siteng",
""
],
[
"Gao",
"Songyang",
""
],
[
"Qin",
"Chengwei",
""
],
[
"Wu",
"Kejian",
""
],
[
"Fan",
"Zhaoxin",
""
],
[
"Qiao",
"Ziyue",
""
],
[
"Wang",
"Donglin",
""
]
] | TITLE: Unicorn: Text-Only Data Synthesis for Vision Language Model Training
ABSTRACT: Training vision-language models (VLMs) typically requires large-scale,
high-quality image-text pairs, but collecting or synthesizing such data is
costly. In contrast, text data is abundant and inexpensive, prompting the
question: can high-quality multimodal training data be synthesized purely from
text? To tackle this, we propose a cross-integrated three-stage multimodal data
synthesis framework, which generates two datasets: Unicorn-1.2M and
Unicorn-471K-Instruction. In Stage 1: Diverse Caption Data Synthesis, we
construct 1.2M semantically diverse high-quality captions by expanding sparse
caption seeds using large language models (LLMs). In Stage 2:
Instruction-Tuning Data Generation, we further process 471K captions into
multi-turn instruction-tuning tasks to support complex reasoning. Finally, in
Stage 3: Modality Representation Transfer, these textual captions
representations are transformed into visual representations, resulting in
diverse synthetic image representations. This three-stage process enables us to
construct Unicorn-1.2M for pretraining and Unicorn-471K-Instruction for
instruction-tuning, without relying on real images. By eliminating the
dependency on real images while maintaining data quality and diversity, our
framework offers a cost-effective and scalable solution for VLMs training. Code
is available at https://github.com/Yu-xm/Unicorn.git.
|
2503.22668 | Sindhu Hegde | Sindhu B Hegde, K R Prajwal, Taein Kwon, Andrew Zisserman | Understanding Co-speech Gestures in-the-wild | Main paper - 11 pages, 4 figures, Supplementary - 5 pages, 4 figures | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Co-speech gestures play a vital role in non-verbal communication. In this
paper, we introduce a new framework for co-speech gesture understanding in the
wild. Specifically, we propose three new tasks and benchmarks to evaluate a
model's capability to comprehend gesture-text-speech associations: (i)
gesture-based retrieval, (ii) gestured word spotting, and (iii) active speaker
detection using gestures. We present a new approach that learns a tri-modal
speech-text-video-gesture representation to solve these tasks. By leveraging a
combination of global phrase contrastive loss and local gesture-word coupling
loss, we demonstrate that a strong gesture representation can be learned in a
weakly supervised manner from videos in the wild. Our learned representations
outperform previous methods, including large vision-language models (VLMs),
across all three tasks. Further analysis reveals that speech and text
modalities capture distinct gesture-related signals, underscoring the
advantages of learning a shared tri-modal embedding space. The dataset, model,
and code are available at: https://www.robots.ox.ac.uk/~vgg/research/jegal
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 17:55:52 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Hegde",
"Sindhu B",
""
],
[
"Prajwal",
"K R",
""
],
[
"Kwon",
"Taein",
""
],
[
"Zisserman",
"Andrew",
""
]
] | TITLE: Understanding Co-speech Gestures in-the-wild
ABSTRACT: Co-speech gestures play a vital role in non-verbal communication. In this
paper, we introduce a new framework for co-speech gesture understanding in the
wild. Specifically, we propose three new tasks and benchmarks to evaluate a
model's capability to comprehend gesture-text-speech associations: (i)
gesture-based retrieval, (ii) gestured word spotting, and (iii) active speaker
detection using gestures. We present a new approach that learns a tri-modal
speech-text-video-gesture representation to solve these tasks. By leveraging a
combination of global phrase contrastive loss and local gesture-word coupling
loss, we demonstrate that a strong gesture representation can be learned in a
weakly supervised manner from videos in the wild. Our learned representations
outperform previous methods, including large vision-language models (VLMs),
across all three tasks. Further analysis reveals that speech and text
modalities capture distinct gesture-related signals, underscoring the
advantages of learning a shared tri-modal embedding space. The dataset, model,
and code are available at: https://www.robots.ox.ac.uk/~vgg/research/jegal
|
2503.22675 | Jiakai Tang | Jiakai Tang, Sunhao Dai, Teng Shi, Jun Xu, Xu Chen, Wen Chen, Wu Jian,
Yuning Jiang | Think Before Recommend: Unleashing the Latent Reasoning Power for
Sequential Recommendation | null | null | null | null | cs.IR cs.AI cs.CL | http://creativecommons.org/licenses/by/4.0/ | Sequential Recommendation (SeqRec) aims to predict the next item by capturing
sequential patterns from users' historical interactions, playing a crucial role
in many real-world recommender systems. However, existing approaches
predominantly adopt a direct forward computation paradigm, where the final
hidden state of the sequence encoder serves as the user representation. We
argue that this inference paradigm, due to its limited computational depth,
struggles to model the complex evolving nature of user preferences and lacks a
nuanced understanding of long-tail items, leading to suboptimal performance. To
address this issue, we propose \textbf{ReaRec}, the first inference-time
computing framework for recommender systems, which enhances user
representations through implicit multi-step reasoning. Specifically, ReaRec
autoregressively feeds the sequence's last hidden state into the sequential
recommender while incorporating special reasoning position embeddings to
decouple the original item encoding space from the multi-step reasoning space.
Moreover, we introduce two lightweight reasoning-based learning methods,
Ensemble Reasoning Learning (ERL) and Progressive Reasoning Learning (PRL), to
further effectively exploit ReaRec's reasoning potential. Extensive experiments
on five public real-world datasets and different SeqRec architectures
demonstrate the generality and effectiveness of our proposed ReaRec.
Remarkably, post-hoc analyses reveal that ReaRec significantly elevates the
performance ceiling of multiple sequential recommendation backbones by
approximately 30\%-50\%. Thus, we believe this work can open a new and
promising avenue for future research in inference-time computing for sequential
recommendation.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 17:59:03 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Tang",
"Jiakai",
""
],
[
"Dai",
"Sunhao",
""
],
[
"Shi",
"Teng",
""
],
[
"Xu",
"Jun",
""
],
[
"Chen",
"Xu",
""
],
[
"Chen",
"Wen",
""
],
[
"Jian",
"Wu",
""
],
[
"Jiang",
"Yuning",
""
]
] | TITLE: Think Before Recommend: Unleashing the Latent Reasoning Power for
Sequential Recommendation
ABSTRACT: Sequential Recommendation (SeqRec) aims to predict the next item by capturing
sequential patterns from users' historical interactions, playing a crucial role
in many real-world recommender systems. However, existing approaches
predominantly adopt a direct forward computation paradigm, where the final
hidden state of the sequence encoder serves as the user representation. We
argue that this inference paradigm, due to its limited computational depth,
struggles to model the complex evolving nature of user preferences and lacks a
nuanced understanding of long-tail items, leading to suboptimal performance. To
address this issue, we propose \textbf{ReaRec}, the first inference-time
computing framework for recommender systems, which enhances user
representations through implicit multi-step reasoning. Specifically, ReaRec
autoregressively feeds the sequence's last hidden state into the sequential
recommender while incorporating special reasoning position embeddings to
decouple the original item encoding space from the multi-step reasoning space.
Moreover, we introduce two lightweight reasoning-based learning methods,
Ensemble Reasoning Learning (ERL) and Progressive Reasoning Learning (PRL), to
further effectively exploit ReaRec's reasoning potential. Extensive experiments
on five public real-world datasets and different SeqRec architectures
demonstrate the generality and effectiveness of our proposed ReaRec.
Remarkably, post-hoc analyses reveal that ReaRec significantly elevates the
performance ceiling of multiple sequential recommendation backbones by
approximately 30\%-50\%. Thus, we believe this work can open a new and
promising avenue for future research in inference-time computing for sequential
recommendation.
|
2503.22677 | Ruining Li | Ruining Li, Chuanxia Zheng, Christian Rupprecht, Andrea Vedaldi | DSO: Aligning 3D Generators with Simulation Feedback for Physical
Soundness | Project page: https://ruiningli.com/dso | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Most 3D object generators focus on aesthetic quality, often neglecting
physical constraints necessary in applications. One such constraint is that the
3D object should be self-supporting, i.e., remains balanced under gravity.
Prior approaches to generating stable 3D objects used differentiable physics
simulators to optimize geometry at test-time, which is slow, unstable, and
prone to local optima. Inspired by the literature on aligning generative models
to external feedback, we propose Direct Simulation Optimization (DSO), a
framework to use the feedback from a (non-differentiable) simulator to increase
the likelihood that the 3D generator outputs stable 3D objects directly. We
construct a dataset of 3D objects labeled with a stability score obtained from
the physics simulator. We can then fine-tune the 3D generator using the
stability score as the alignment metric, via direct preference optimization
(DPO) or direct reward optimization (DRO), a novel objective, which we
introduce, to align diffusion models without requiring pairwise preferences.
Our experiments show that the fine-tuned feed-forward generator, using either
DPO or DRO objective, is much faster and more likely to produce stable objects
than test-time optimization. Notably, the DSO framework works even without any
ground-truth 3D objects for training, allowing the 3D generator to self-improve
by automatically collecting simulation feedback on its own outputs.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 17:59:53 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Li",
"Ruining",
""
],
[
"Zheng",
"Chuanxia",
""
],
[
"Rupprecht",
"Christian",
""
],
[
"Vedaldi",
"Andrea",
""
]
] | TITLE: DSO: Aligning 3D Generators with Simulation Feedback for Physical
Soundness
ABSTRACT: Most 3D object generators focus on aesthetic quality, often neglecting
physical constraints necessary in applications. One such constraint is that the
3D object should be self-supporting, i.e., remains balanced under gravity.
Prior approaches to generating stable 3D objects used differentiable physics
simulators to optimize geometry at test-time, which is slow, unstable, and
prone to local optima. Inspired by the literature on aligning generative models
to external feedback, we propose Direct Simulation Optimization (DSO), a
framework to use the feedback from a (non-differentiable) simulator to increase
the likelihood that the 3D generator outputs stable 3D objects directly. We
construct a dataset of 3D objects labeled with a stability score obtained from
the physics simulator. We can then fine-tune the 3D generator using the
stability score as the alignment metric, via direct preference optimization
(DPO) or direct reward optimization (DRO), a novel objective, which we
introduce, to align diffusion models without requiring pairwise preferences.
Our experiments show that the fine-tuned feed-forward generator, using either
DPO or DRO objective, is much faster and more likely to produce stable objects
than test-time optimization. Notably, the DSO framework works even without any
ground-truth 3D objects for training, allowing the 3D generator to self-improve
by automatically collecting simulation feedback on its own outputs.
|
2503.22679 | Weiqi Li | Weiqi Li, Xuanyu Zhang, Shijie Zhao, Yabin Zhang, Junlin Li, Li Zhang,
Jian Zhang | Q-Insight: Understanding Image Quality via Visual Reinforcement Learning | Technical report | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Image quality assessment (IQA) focuses on the perceptual visual quality of
images, playing a crucial role in downstream tasks such as image
reconstruction, compression, and generation. The rapid advancement of
multi-modal large language models (MLLMs) has significantly broadened the scope
of IQA, moving toward comprehensive image quality understanding that
incorporates content analysis, degradation perception, and comparison reasoning
beyond mere numerical scoring. Previous MLLM-based methods typically either
generate numerical scores lacking interpretability or heavily rely on
supervised fine-tuning (SFT) using large-scale annotated datasets to provide
descriptive assessments, limiting their flexibility and applicability. In this
paper, we propose Q-Insight, a reinforcement learning-based model built upon
group relative policy optimization (GRPO), which demonstrates strong visual
reasoning capability for image quality understanding while requiring only a
limited amount of rating scores and degradation labels. By jointly optimizing
score regression and degradation perception tasks with carefully designed
reward functions, our approach effectively exploits their mutual benefits for
enhanced performance. Extensive experiments demonstrate that Q-Insight
substantially outperforms existing state-of-the-art methods in both score
regression and degradation perception tasks, while exhibiting impressive
zero-shot generalization to comparison reasoning tasks. Code will be available
at https://github.com/lwq20020127/Q-Insight.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 17:59:54 GMT"
}
] | 2025-03-31T00:00:00 | [
[
"Li",
"Weiqi",
""
],
[
"Zhang",
"Xuanyu",
""
],
[
"Zhao",
"Shijie",
""
],
[
"Zhang",
"Yabin",
""
],
[
"Li",
"Junlin",
""
],
[
"Zhang",
"Li",
""
],
[
"Zhang",
"Jian",
""
]
] | TITLE: Q-Insight: Understanding Image Quality via Visual Reinforcement Learning
ABSTRACT: Image quality assessment (IQA) focuses on the perceptual visual quality of
images, playing a crucial role in downstream tasks such as image
reconstruction, compression, and generation. The rapid advancement of
multi-modal large language models (MLLMs) has significantly broadened the scope
of IQA, moving toward comprehensive image quality understanding that
incorporates content analysis, degradation perception, and comparison reasoning
beyond mere numerical scoring. Previous MLLM-based methods typically either
generate numerical scores lacking interpretability or heavily rely on
supervised fine-tuning (SFT) using large-scale annotated datasets to provide
descriptive assessments, limiting their flexibility and applicability. In this
paper, we propose Q-Insight, a reinforcement learning-based model built upon
group relative policy optimization (GRPO), which demonstrates strong visual
reasoning capability for image quality understanding while requiring only a
limited amount of rating scores and degradation labels. By jointly optimizing
score regression and degradation perception tasks with carefully designed
reward functions, our approach effectively exploits their mutual benefits for
enhanced performance. Extensive experiments demonstrate that Q-Insight
substantially outperforms existing state-of-the-art methods in both score
regression and degradation perception tasks, while exhibiting impressive
zero-shot generalization to comparison reasoning tasks. Code will be available
at https://github.com/lwq20020127/Q-Insight.
|
2503.19588 | Mia Siemon | Mia Siemon, Ivan Nikolov, Thomas B. Moeslund and Kamal Nasrollahi | Video Anomaly Detection with Contours -- A Study | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In Pose-based Video Anomaly Detection prior art is rooted on the assumption
that abnormal events can be mostly regarded as a result of uncommon human
behavior. Opposed to utilizing skeleton representations of humans, however, we
investigate the potential of learning recurrent motion patterns of normal human
behavior using 2D contours. Keeping all advantages of pose-based methods, such
as increased object anonymization, the shift from human skeletons to contours
is hypothesized to leave the opportunity to cover more object categories open
for future research. We propose formulating the problem as a regression and a
classification task, and additionally explore two distinct data representation
techniques for contours. To further reduce the computational complexity of
Pose-based Video Anomaly Detection solutions, all methods in this study are
based on shallow Neural Networks from the field of Deep Learning, and evaluated
on the three most prominent benchmark datasets within Video Anomaly Detection
and their human-related counterparts, totaling six datasets. Our results
indicate that this novel perspective on Pose-based Video Anomaly Detection
marks a promising direction for future research.
| [
{
"version": "v1",
"created": "Tue, 25 Mar 2025 12:11:50 GMT"
}
] | 2025-03-30T00:00:00 | [
[
"Siemon",
"Mia",
""
],
[
"Nikolov",
"Ivan",
""
],
[
"Moeslund",
"Thomas B.",
""
],
[
"Nasrollahi",
"Kamal",
""
]
] | TITLE: Video Anomaly Detection with Contours -- A Study
ABSTRACT: In Pose-based Video Anomaly Detection prior art is rooted on the assumption
that abnormal events can be mostly regarded as a result of uncommon human
behavior. Opposed to utilizing skeleton representations of humans, however, we
investigate the potential of learning recurrent motion patterns of normal human
behavior using 2D contours. Keeping all advantages of pose-based methods, such
as increased object anonymization, the shift from human skeletons to contours
is hypothesized to leave the opportunity to cover more object categories open
for future research. We propose formulating the problem as a regression and a
classification task, and additionally explore two distinct data representation
techniques for contours. To further reduce the computational complexity of
Pose-based Video Anomaly Detection solutions, all methods in this study are
based on shallow Neural Networks from the field of Deep Learning, and evaluated
on the three most prominent benchmark datasets within Video Anomaly Detection
and their human-related counterparts, totaling six datasets. Our results
indicate that this novel perspective on Pose-based Video Anomaly Detection
marks a promising direction for future research.
|
2503.19670 | Saurav Sharma | Saurav Sharma, Didier Mutter, Nicolas Padoy | fine-CLIP: Enhancing Zero-Shot Fine-Grained Surgical Action Recognition
with Vision-Language Models | 6 pages, 3 tables, 3 figures | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | While vision-language models like CLIP have advanced zero-shot surgical phase
recognition, they struggle with fine-grained surgical activities, especially
action triplets. This limitation arises because current CLIP formulations rely
on global image features, which overlook the fine-grained semantics and
contextual details crucial for complex tasks like zero-shot triplet
recognition. Furthermore, these models do not explore the hierarchical
structure inherent in triplets, reducing their ability to generalize to novel
triplets. To address these challenges, we propose fine-CLIP, which learns
object-centric features and leverages the hierarchy in triplet formulation. Our
approach integrates three components: hierarchical prompt modeling to capture
shared semantics, LoRA-based vision backbone adaptation for enhanced feature
extraction, and a graph-based condensation strategy that groups similar patch
features into meaningful object clusters. Since triplet classification is a
challenging task, we introduce an alternative yet meaningful base-to-novel
generalization benchmark with two settings on the CholecT50 dataset:
Unseen-Target, assessing adaptability to triplets with novel anatomical
structures, and Unseen-Instrument-Verb, where models need to generalize to
novel instrument-verb interactions. fine-CLIP shows significant improvements in
F1 and mAP, enhancing zero-shot recognition of novel surgical triplets.
| [
{
"version": "v1",
"created": "Tue, 25 Mar 2025 13:57:02 GMT"
}
] | 2025-03-30T00:00:00 | [
[
"Sharma",
"Saurav",
""
],
[
"Mutter",
"Didier",
""
],
[
"Padoy",
"Nicolas",
""
]
] | TITLE: fine-CLIP: Enhancing Zero-Shot Fine-Grained Surgical Action Recognition
with Vision-Language Models
ABSTRACT: While vision-language models like CLIP have advanced zero-shot surgical phase
recognition, they struggle with fine-grained surgical activities, especially
action triplets. This limitation arises because current CLIP formulations rely
on global image features, which overlook the fine-grained semantics and
contextual details crucial for complex tasks like zero-shot triplet
recognition. Furthermore, these models do not explore the hierarchical
structure inherent in triplets, reducing their ability to generalize to novel
triplets. To address these challenges, we propose fine-CLIP, which learns
object-centric features and leverages the hierarchy in triplet formulation. Our
approach integrates three components: hierarchical prompt modeling to capture
shared semantics, LoRA-based vision backbone adaptation for enhanced feature
extraction, and a graph-based condensation strategy that groups similar patch
features into meaningful object clusters. Since triplet classification is a
challenging task, we introduce an alternative yet meaningful base-to-novel
generalization benchmark with two settings on the CholecT50 dataset:
Unseen-Target, assessing adaptability to triplets with novel anatomical
structures, and Unseen-Instrument-Verb, where models need to generalize to
novel instrument-verb interactions. fine-CLIP shows significant improvements in
F1 and mAP, enhancing zero-shot recognition of novel surgical triplets.
|
2503.19860 | Junzhi Ning | Junzhi Ning, Dominic Marshall, Yijian Gao, Xiaodan Xing Yang Nan,
Yingying Fang, Sheng Zhang, Matthieu Komorowski, Guang Yang | Unpaired Translation of Chest X-ray Images for Lung Opacity Diagnosis
via Adaptive Activation Masks and Cross-Domain Alignment | null | null | null | null | eess.IV cs.CV | http://creativecommons.org/licenses/by/4.0/ | Chest X-ray radiographs (CXRs) play a pivotal role in diagnosing and
monitoring cardiopulmonary diseases. However, lung opacities in CXRs frequently
obscure anatomical structures, impeding clear identification of lung borders
and complicating the localization of pathology. This challenge significantly
hampers segmentation accuracy and precise lesion identification, which are
crucial for diagnosis. To tackle these issues, our study proposes an unpaired
CXR translation framework that converts CXRs with lung opacities into
counterparts without lung opacities while preserving semantic features. Central
to our approach is the use of adaptive activation masks to selectively modify
opacity regions in lung CXRs. Cross-domain alignment ensures translated CXRs
without opacity issues align with feature maps and prediction labels from a
pre-trained CXR lesion classifier, facilitating the interpretability of the
translation process. We validate our method using RSNA, MIMIC-CXR-JPG and JSRT
datasets, demonstrating superior translation quality through lower Frechet
Inception Distance (FID) and Kernel Inception Distance (KID) scores compared to
existing methods (FID: 67.18 vs. 210.4, KID: 0.01604 vs. 0.225). Evaluation on
RSNA opacity, MIMIC acute respiratory distress syndrome (ARDS) patient CXRs and
JSRT CXRs show our method enhances segmentation accuracy of lung borders and
improves lesion classification, further underscoring its potential in clinical
settings (RSNA: mIoU: 76.58% vs. 62.58%, Sensitivity: 85.58% vs. 77.03%; MIMIC
ARDS: mIoU: 86.20% vs. 72.07%, Sensitivity: 92.68% vs. 86.85%; JSRT: mIoU:
91.08% vs. 85.6%, Sensitivity: 97.62% vs. 95.04%). Our approach advances CXR
imaging analysis, especially in investigating segmentation impacts through
image translation techniques.
| [
{
"version": "v1",
"created": "Tue, 25 Mar 2025 17:26:17 GMT"
}
] | 2025-03-30T00:00:00 | [
[
"Ning",
"Junzhi",
""
],
[
"Marshall",
"Dominic",
""
],
[
"Gao",
"Yijian",
""
],
[
"Nan",
"Xiaodan Xing Yang",
""
],
[
"Fang",
"Yingying",
""
],
[
"Zhang",
"Sheng",
""
],
[
"Komorowski",
"Matthieu",
""
],
[
"Yang",
"Guang",
""
]
] | TITLE: Unpaired Translation of Chest X-ray Images for Lung Opacity Diagnosis
via Adaptive Activation Masks and Cross-Domain Alignment
ABSTRACT: Chest X-ray radiographs (CXRs) play a pivotal role in diagnosing and
monitoring cardiopulmonary diseases. However, lung opacities in CXRs frequently
obscure anatomical structures, impeding clear identification of lung borders
and complicating the localization of pathology. This challenge significantly
hampers segmentation accuracy and precise lesion identification, which are
crucial for diagnosis. To tackle these issues, our study proposes an unpaired
CXR translation framework that converts CXRs with lung opacities into
counterparts without lung opacities while preserving semantic features. Central
to our approach is the use of adaptive activation masks to selectively modify
opacity regions in lung CXRs. Cross-domain alignment ensures translated CXRs
without opacity issues align with feature maps and prediction labels from a
pre-trained CXR lesion classifier, facilitating the interpretability of the
translation process. We validate our method using RSNA, MIMIC-CXR-JPG and JSRT
datasets, demonstrating superior translation quality through lower Frechet
Inception Distance (FID) and Kernel Inception Distance (KID) scores compared to
existing methods (FID: 67.18 vs. 210.4, KID: 0.01604 vs. 0.225). Evaluation on
RSNA opacity, MIMIC acute respiratory distress syndrome (ARDS) patient CXRs and
JSRT CXRs show our method enhances segmentation accuracy of lung borders and
improves lesion classification, further underscoring its potential in clinical
settings (RSNA: mIoU: 76.58% vs. 62.58%, Sensitivity: 85.58% vs. 77.03%; MIMIC
ARDS: mIoU: 86.20% vs. 72.07%, Sensitivity: 92.68% vs. 86.85%; JSRT: mIoU:
91.08% vs. 85.6%, Sensitivity: 97.62% vs. 95.04%). Our approach advances CXR
imaging analysis, especially in investigating segmentation impacts through
image translation techniques.
|
2012.04726 | Jeff Da | Jeff Da and Maxwell Forbes and Rowan Zellers and Anthony Zheng and
Jena D. Hwang and Antoine Bosselut and Yejin Choi | Edited Media Understanding Frames: Reasoning About the Intent and
Implications of Visual Misinformation | ACL 2021 | null | null | null | cs.CL cs.CV | http://creativecommons.org/licenses/by/4.0/ | Multimodal disinformation, from 'deepfakes' to simple edits that deceive, is
an important societal problem. Yet at the same time, the vast majority of media
edits are harmless -- such as a filtered vacation photo. The difference between
this example, and harmful edits that spread disinformation, is one of intent.
Recognizing and describing this intent is a major challenge for today's AI
systems.
We present the task of Edited Media Understanding, requiring models to answer
open-ended questions that capture the intent and implications of an image edit.
We introduce a dataset for our task, EMU, with 48k question-answer pairs
written in rich natural language. We evaluate a wide variety of
vision-and-language models for our task, and introduce a new model PELICAN,
which builds upon recent progress in pretrained multimodal representations. Our
model obtains promising results on our dataset, with humans rating its answers
as accurate 40.35% of the time. At the same time, there is still much work to
be done -- humans prefer human-annotated captions 93.56% of the time -- and we
provide analysis that highlights areas for further progress.
| [
{
"version": "v1",
"created": "Tue, 8 Dec 2020 20:30:43 GMT"
},
{
"version": "v2",
"created": "Wed, 26 Mar 2025 20:17:54 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Da",
"Jeff",
""
],
[
"Forbes",
"Maxwell",
""
],
[
"Zellers",
"Rowan",
""
],
[
"Zheng",
"Anthony",
""
],
[
"Hwang",
"Jena D.",
""
],
[
"Bosselut",
"Antoine",
""
],
[
"Choi",
"Yejin",
""
]
] | TITLE: Edited Media Understanding Frames: Reasoning About the Intent and
Implications of Visual Misinformation
ABSTRACT: Multimodal disinformation, from 'deepfakes' to simple edits that deceive, is
an important societal problem. Yet at the same time, the vast majority of media
edits are harmless -- such as a filtered vacation photo. The difference between
this example, and harmful edits that spread disinformation, is one of intent.
Recognizing and describing this intent is a major challenge for today's AI
systems.
We present the task of Edited Media Understanding, requiring models to answer
open-ended questions that capture the intent and implications of an image edit.
We introduce a dataset for our task, EMU, with 48k question-answer pairs
written in rich natural language. We evaluate a wide variety of
vision-and-language models for our task, and introduce a new model PELICAN,
which builds upon recent progress in pretrained multimodal representations. Our
model obtains promising results on our dataset, with humans rating its answers
as accurate 40.35% of the time. At the same time, there is still much work to
be done -- humans prefer human-annotated captions 93.56% of the time -- and we
provide analysis that highlights areas for further progress.
|
2301.11923 | Alexej Schelle Dr. | A. Schelle and H. L\"uling | Information loss from dimensionality reduction in 5D-Gaussian spectral
data | 4 pages, 3 figures | Whitepaper on arXiv.org (2023) | null | null | physics.data-an cs.LG quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Understanding the loss of information in spectral analytics is a crucial
first step towards finding root causes for failures and uncertainties using
spectral data in artificial intelligence models built from modern complex data
science applications. Here, we show from an elementary Shannon entropy model
analysis with quantum statistics of Gaussian distributed spectral data, that
the relative loss of information from dimensionality reduction due to the
projection of an initial five-dimensional dataset onto two-dimensional diagrams
is less than one percent in the parameter range of small data sets with sample
sizes on the order of few hundred data samples. From our analysis, we also
conclude that the density and expectation value of the entropy probability
distribution increases with the sample number and sample size using artificial
data models derived from random sampling Monte Carlo simulation methods.
| [
{
"version": "v1",
"created": "Sun, 22 Jan 2023 14:51:35 GMT"
},
{
"version": "v2",
"created": "Sat, 23 Dec 2023 12:56:33 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Schelle",
"A.",
""
],
[
"Lüling",
"H.",
""
]
] | TITLE: Information loss from dimensionality reduction in 5D-Gaussian spectral
data
ABSTRACT: Understanding the loss of information in spectral analytics is a crucial
first step towards finding root causes for failures and uncertainties using
spectral data in artificial intelligence models built from modern complex data
science applications. Here, we show from an elementary Shannon entropy model
analysis with quantum statistics of Gaussian distributed spectral data, that
the relative loss of information from dimensionality reduction due to the
projection of an initial five-dimensional dataset onto two-dimensional diagrams
is less than one percent in the parameter range of small data sets with sample
sizes on the order of few hundred data samples. From our analysis, we also
conclude that the density and expectation value of the entropy probability
distribution increases with the sample number and sample size using artificial
data models derived from random sampling Monte Carlo simulation methods.
|
2308.07421 | Hamidreza Behjoo | Hamidreza Behjoo, Michael Chertkov | U-Turn Diffusion | null | Entropy 2025 | 10.3390/e27040343 | null | cs.LG cs.CV | http://creativecommons.org/licenses/by/4.0/ | We investigate diffusion models generating synthetic samples from the
probability distribution represented by the Ground Truth (GT) samples. We focus
on how GT sample information is encoded in the Score Function (SF), computed
(not simulated) from the Wiener-Ito (WI) linear forward process in the
artifical time $t\in [0\to \infty]$, and then used as a nonlinear drift in the
simulated WI reverse process with $t\in [\infty\to 0]$. We propose U-Turn
diffusion, an augmentation of a pre-trained diffusion model, which shortens the
forward and reverse processes to $t\in [0\to T_u]$ and $t\in [T_u\to 0]$. The
U-Turn reverse process is initialized at $T_u$ with a sample from the
probability distribution of the forward process (initialized at $t=0$ with a GT
sample) ensuring a detailed balance relation between the shorten forward and
reverse processes. Our experiments on the class-conditioned SF of the ImageNet
dataset and the multi-class, single SF of the CIFAR-10 dataset reveal a
critical Memorization Time $ T_m $, beyond which generated samples diverge from
the GT sample used to initialize the U-Turn scheme, and a Speciation Time $ T_s
$, where for $ T_u > T_s > T_m $, samples begin representing different classes.
We further examine the role of SF non-linearity through a Gaussian Test,
comparing empirical and Gaussian-approximated U-Turn auto-correlation
functions, and showing that the SF becomes effectively affine for $ t > T_s $,
and approximately affine for $t\in [T_m,T_s]$.
| [
{
"version": "v1",
"created": "Mon, 14 Aug 2023 19:21:28 GMT"
},
{
"version": "v2",
"created": "Wed, 22 May 2024 20:00:17 GMT"
},
{
"version": "v3",
"created": "Wed, 25 Dec 2024 18:35:24 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Behjoo",
"Hamidreza",
""
],
[
"Chertkov",
"Michael",
""
]
] | TITLE: U-Turn Diffusion
ABSTRACT: We investigate diffusion models generating synthetic samples from the
probability distribution represented by the Ground Truth (GT) samples. We focus
on how GT sample information is encoded in the Score Function (SF), computed
(not simulated) from the Wiener-Ito (WI) linear forward process in the
artifical time $t\in [0\to \infty]$, and then used as a nonlinear drift in the
simulated WI reverse process with $t\in [\infty\to 0]$. We propose U-Turn
diffusion, an augmentation of a pre-trained diffusion model, which shortens the
forward and reverse processes to $t\in [0\to T_u]$ and $t\in [T_u\to 0]$. The
U-Turn reverse process is initialized at $T_u$ with a sample from the
probability distribution of the forward process (initialized at $t=0$ with a GT
sample) ensuring a detailed balance relation between the shorten forward and
reverse processes. Our experiments on the class-conditioned SF of the ImageNet
dataset and the multi-class, single SF of the CIFAR-10 dataset reveal a
critical Memorization Time $ T_m $, beyond which generated samples diverge from
the GT sample used to initialize the U-Turn scheme, and a Speciation Time $ T_s
$, where for $ T_u > T_s > T_m $, samples begin representing different classes.
We further examine the role of SF non-linearity through a Gaussian Test,
comparing empirical and Gaussian-approximated U-Turn auto-correlation
functions, and showing that the SF becomes effectively affine for $ t > T_s $,
and approximately affine for $t\in [T_m,T_s]$.
|
2310.04722 | Monan Zhou Dr | Monan Zhou, Shangda Wu, Shaohua Ji, Zijin Li, Wei Li | A Holistic Evaluation of Piano Sound Quality | 15 pages, 9 figures | Proceedings of the 10th Conference on Sound and Music Technology.
CSMT 2023. Lecture Notes in Electrical Engineering, vol 1268. Springer,
Singapore | 10.1007/978-981-97-7962-8_1 | 23638935599966770924 | cs.SD cs.AI eess.AS | http://creativecommons.org/licenses/by/4.0/ | This paper aims to develop a holistic evaluation method for piano sound
quality to assist in purchasing decisions. Unlike previous studies that focused
on the effect of piano performance techniques on sound quality, this study
evaluates the inherent sound quality of different pianos. To derive quality
evaluation systems, the study uses subjective questionnaires based on a piano
sound quality dataset. The method selects the optimal piano classification
models by comparing the fine-tuning results of different pre-training models of
Convolutional Neural Networks (CNN). To improve the interpretability of the
models, the study applies Equivalent Rectangular Bandwidth (ERB) analysis. The
results reveal that musically trained individuals are better able to
distinguish between the sound quality differences of different pianos. The best
fine-tuned CNN pre-trained backbone achieves a high accuracy of 98.3% as the
piano classifier. However, the dataset is limited, and the audio is sliced to
increase its quantity, resulting in a lack of diversity and balance, so we use
focal loss to reduce the impact of data imbalance. To optimize the method, the
dataset will be expanded, or few-shot learning techniques will be employed in
future research.
| [
{
"version": "v1",
"created": "Sat, 7 Oct 2023 07:51:34 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 02:31:56 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Zhou",
"Monan",
""
],
[
"Wu",
"Shangda",
""
],
[
"Ji",
"Shaohua",
""
],
[
"Li",
"Zijin",
""
],
[
"Li",
"Wei",
""
]
] | TITLE: A Holistic Evaluation of Piano Sound Quality
ABSTRACT: This paper aims to develop a holistic evaluation method for piano sound
quality to assist in purchasing decisions. Unlike previous studies that focused
on the effect of piano performance techniques on sound quality, this study
evaluates the inherent sound quality of different pianos. To derive quality
evaluation systems, the study uses subjective questionnaires based on a piano
sound quality dataset. The method selects the optimal piano classification
models by comparing the fine-tuning results of different pre-training models of
Convolutional Neural Networks (CNN). To improve the interpretability of the
models, the study applies Equivalent Rectangular Bandwidth (ERB) analysis. The
results reveal that musically trained individuals are better able to
distinguish between the sound quality differences of different pianos. The best
fine-tuned CNN pre-trained backbone achieves a high accuracy of 98.3% as the
piano classifier. However, the dataset is limited, and the audio is sliced to
increase its quantity, resulting in a lack of diversity and balance, so we use
focal loss to reduce the impact of data imbalance. To optimize the method, the
dataset will be expanded, or few-shot learning techniques will be employed in
future research.
|
2311.15917 | Zhanbo Liang | Zhanbo Liang, Jie Guo, Weidong Qiu, Zheng Huang and Shujun Li | When Graph Convolution Meets Double Attention: Online Privacy Disclosure
Detection with Multi-Label Text Classification | The manuscript is accepted by Data Mining and Knowledge
Discovery(ECML PKDD Journal track) | null | 10.1007/s10618-023-00992-y | null | cs.CR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With the rise of Web 2.0 platforms such as online social media, people's
private information, such as their location, occupation and even family
information, is often inadvertently disclosed through online discussions.
Therefore, it is important to detect such unwanted privacy disclosures to help
alert people affected and the online platform. In this paper, privacy
disclosure detection is modeled as a multi-label text classification (MLTC)
problem, and a new privacy disclosure detection model is proposed to construct
an MLTC classifier for detecting online privacy disclosures. This classifier
takes an online post as the input and outputs multiple labels, each reflecting
a possible privacy disclosure. The proposed presentation method combines three
different sources of information, the input text itself, the label-to-text
correlation and the label-to-label correlation. A double-attention mechanism is
used to combine the first two sources of information, and a graph convolutional
network (GCN) is employed to extract the third source of information that is
then used to help fuse features extracted from the first two sources of
information. Our extensive experimental results, obtained on a public dataset
of privacy-disclosing posts on Twitter, demonstrated that our proposed privacy
disclosure detection method significantly and consistently outperformed other
state-of-the-art methods in terms of all key performance indicators.
| [
{
"version": "v1",
"created": "Mon, 27 Nov 2023 15:25:17 GMT"
},
{
"version": "v2",
"created": "Wed, 20 Dec 2023 08:40:33 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Liang",
"Zhanbo",
""
],
[
"Guo",
"Jie",
""
],
[
"Qiu",
"Weidong",
""
],
[
"Huang",
"Zheng",
""
],
[
"Li",
"Shujun",
""
]
] | TITLE: When Graph Convolution Meets Double Attention: Online Privacy Disclosure
Detection with Multi-Label Text Classification
ABSTRACT: With the rise of Web 2.0 platforms such as online social media, people's
private information, such as their location, occupation and even family
information, is often inadvertently disclosed through online discussions.
Therefore, it is important to detect such unwanted privacy disclosures to help
alert people affected and the online platform. In this paper, privacy
disclosure detection is modeled as a multi-label text classification (MLTC)
problem, and a new privacy disclosure detection model is proposed to construct
an MLTC classifier for detecting online privacy disclosures. This classifier
takes an online post as the input and outputs multiple labels, each reflecting
a possible privacy disclosure. The proposed presentation method combines three
different sources of information, the input text itself, the label-to-text
correlation and the label-to-label correlation. A double-attention mechanism is
used to combine the first two sources of information, and a graph convolutional
network (GCN) is employed to extract the third source of information that is
then used to help fuse features extracted from the first two sources of
information. Our extensive experimental results, obtained on a public dataset
of privacy-disclosing posts on Twitter, demonstrated that our proposed privacy
disclosure detection method significantly and consistently outperformed other
state-of-the-art methods in terms of all key performance indicators.
|
2311.16909 | Hylke Donker | H. C. Donker, D. Neijzen, J. de Jong, G. A. Lunter | Multinomial belief networks for healthcare data | 18 pages, 4 figs; supplement: 22 pages | PMLR 252, 1-22, 2024 | null | null | stat.ML cs.LG stat.AP | http://creativecommons.org/licenses/by/4.0/ | Healthcare data from patient or population cohorts are often characterized by
sparsity, high missingness and relatively small sample sizes. In addition,
being able to quantify uncertainty is often important in a medical context. To
address these analytical requirements we propose a deep generative Bayesian
model for multinomial count data. We develop a collapsed Gibbs sampling
procedure that takes advantage of a series of augmentation relations, inspired
by the Zhou$\unicode{x2013}$Cong$\unicode{x2013}$Chen model. We visualise the
model's ability to identify coherent substructures in the data using a dataset
of handwritten digits. We then apply it to a large experimental dataset of DNA
mutations in cancer and show that we can identify biologically meaningful
clusters of mutational signatures in a fully data-driven way.
| [
{
"version": "v1",
"created": "Tue, 28 Nov 2023 16:12:50 GMT"
},
{
"version": "v2",
"created": "Mon, 18 Mar 2024 11:53:00 GMT"
},
{
"version": "v3",
"created": "Sat, 6 Apr 2024 11:38:31 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Donker",
"H. C.",
""
],
[
"Neijzen",
"D.",
""
],
[
"de Jong",
"J.",
""
],
[
"Lunter",
"G. A.",
""
]
] | TITLE: Multinomial belief networks for healthcare data
ABSTRACT: Healthcare data from patient or population cohorts are often characterized by
sparsity, high missingness and relatively small sample sizes. In addition,
being able to quantify uncertainty is often important in a medical context. To
address these analytical requirements we propose a deep generative Bayesian
model for multinomial count data. We develop a collapsed Gibbs sampling
procedure that takes advantage of a series of augmentation relations, inspired
by the Zhou$\unicode{x2013}$Cong$\unicode{x2013}$Chen model. We visualise the
model's ability to identify coherent substructures in the data using a dataset
of handwritten digits. We then apply it to a large experimental dataset of DNA
mutations in cancer and show that we can identify biologically meaningful
clusters of mutational signatures in a fully data-driven way.
|
2312.00206 | Haolin Xiong | Haolin Xiong and Sairisheek Muttukuru and Rishi Upadhyay and Pradyumna
Chari and Achuta Kadambi | SparseGS: Real-Time 360{\deg} Sparse View Synthesis using Gaussian
Splatting | Version accepted to 3DV 2025. Project page:
https://github.com/ForMyCat/SparseGS | null | null | null | cs.CV cs.LG eess.IV | http://creativecommons.org/licenses/by/4.0/ | 3D Gaussian Splatting (3DGS) has recently enabled real-time rendering of
unbounded 3D scenes for novel view synthesis. However, this technique requires
dense training views to accurately reconstruct 3D geometry. A limited number of
input views will significantly degrade reconstruction quality, resulting in
artifacts such as "floaters" and "background collapse" at unseen viewpoints. In
this work, we introduce SparseGS, an efficient training pipeline designed to
address the limitations of 3DGS in scenarios with sparse training views.
SparseGS incorporates depth priors, novel depth rendering techniques, and a
pruning heuristic to mitigate floater artifacts, alongside an Unseen Viewpoint
Regularization module to alleviate background collapses. Our extensive
evaluations on the Mip-NeRF360, LLFF, and DTU datasets demonstrate that
SparseGS achieves high-quality reconstruction in both unbounded and
forward-facing scenarios, with as few as 12 and 3 input images, respectively,
while maintaining fast training and real-time rendering capabilities.
| [
{
"version": "v1",
"created": "Thu, 30 Nov 2023 21:38:22 GMT"
},
{
"version": "v2",
"created": "Mon, 13 May 2024 05:11:37 GMT"
},
{
"version": "v3",
"created": "Wed, 26 Mar 2025 19:59:58 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Xiong",
"Haolin",
""
],
[
"Muttukuru",
"Sairisheek",
""
],
[
"Upadhyay",
"Rishi",
""
],
[
"Chari",
"Pradyumna",
""
],
[
"Kadambi",
"Achuta",
""
]
] | TITLE: SparseGS: Real-Time 360{\deg} Sparse View Synthesis using Gaussian
Splatting
ABSTRACT: 3D Gaussian Splatting (3DGS) has recently enabled real-time rendering of
unbounded 3D scenes for novel view synthesis. However, this technique requires
dense training views to accurately reconstruct 3D geometry. A limited number of
input views will significantly degrade reconstruction quality, resulting in
artifacts such as "floaters" and "background collapse" at unseen viewpoints. In
this work, we introduce SparseGS, an efficient training pipeline designed to
address the limitations of 3DGS in scenarios with sparse training views.
SparseGS incorporates depth priors, novel depth rendering techniques, and a
pruning heuristic to mitigate floater artifacts, alongside an Unseen Viewpoint
Regularization module to alleviate background collapses. Our extensive
evaluations on the Mip-NeRF360, LLFF, and DTU datasets demonstrate that
SparseGS achieves high-quality reconstruction in both unbounded and
forward-facing scenarios, with as few as 12 and 3 input images, respectively,
while maintaining fast training and real-time rendering capabilities.
|
2312.07669 | Yibo Xia | Yibo Xia, Lizhen Wang, Xiang Deng, Xiaoyan Luo, Yunhong Wang and Yebin
Liu | GMTalker: Gaussian Mixture-based Audio-Driven Emotional Talking Video
Portraits | Project page: https://bob35buaa.github.io/GMTalker. This work has
been submitted to the IEEE journal for possible publication | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Synthesizing high-fidelity and emotion-controllable talking video portraits,
with audio-lip sync, vivid expressions, realistic head poses, and eye blinks,
has been an important and challenging task in recent years. Most existing
methods suffer in achieving personalized and precise emotion control, smooth
transitions between different emotion states, and the generation of diverse
motions. To tackle these challenges, we present GMTalker, a Gaussian
mixture-based emotional talking portraits generation framework. Specifically,
we propose a Gaussian mixture-based expression generator that can construct a
continuous and disentangled latent space, achieving more flexible emotion
manipulation. Furthermore, we introduce a normalizing flow-based motion
generator pretrained on a large dataset with a wide-range motion to generate
diverse head poses, blinks, and eyeball movements. Finally, we propose a
personalized emotion-guided head generator with an emotion mapping network that
can synthesize high-fidelity and faithful emotional video portraits. Both
quantitative and qualitative experiments demonstrate our method outperforms
previous methods in image quality, photo-realism, emotion accuracy, and motion
diversity.
| [
{
"version": "v1",
"created": "Tue, 12 Dec 2023 19:03:04 GMT"
},
{
"version": "v2",
"created": "Tue, 28 May 2024 17:01:00 GMT"
},
{
"version": "v3",
"created": "Thu, 27 Mar 2025 08:47:12 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Xia",
"Yibo",
""
],
[
"Wang",
"Lizhen",
""
],
[
"Deng",
"Xiang",
""
],
[
"Luo",
"Xiaoyan",
""
],
[
"Wang",
"Yunhong",
""
],
[
"Liu",
"Yebin",
""
]
] | TITLE: GMTalker: Gaussian Mixture-based Audio-Driven Emotional Talking Video
Portraits
ABSTRACT: Synthesizing high-fidelity and emotion-controllable talking video portraits,
with audio-lip sync, vivid expressions, realistic head poses, and eye blinks,
has been an important and challenging task in recent years. Most existing
methods suffer in achieving personalized and precise emotion control, smooth
transitions between different emotion states, and the generation of diverse
motions. To tackle these challenges, we present GMTalker, a Gaussian
mixture-based emotional talking portraits generation framework. Specifically,
we propose a Gaussian mixture-based expression generator that can construct a
continuous and disentangled latent space, achieving more flexible emotion
manipulation. Furthermore, we introduce a normalizing flow-based motion
generator pretrained on a large dataset with a wide-range motion to generate
diverse head poses, blinks, and eyeball movements. Finally, we propose a
personalized emotion-guided head generator with an emotion mapping network that
can synthesize high-fidelity and faithful emotional video portraits. Both
quantitative and qualitative experiments demonstrate our method outperforms
previous methods in image quality, photo-realism, emotion accuracy, and motion
diversity.
|
2401.13174 | Dong Zhang | Dong Zhang, Pingcheng Dong, Long Chen, Kwang-Ting Cheng | Towards Complementary Knowledge Distillation for Efficient Dense Image
Prediction | under submission | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | It has been revealed that small efficient dense image prediction (EDIP)
models, trained using the knowledge distillation (KD) framework, encounter two
key challenges, including maintaining boundary region completeness and
preserving target region connectivity, despite their favorable capacity to
recognize main object regions. In this work, we propose a complementary
boundary and context distillation (BCD) method within the KD framework for
EDIPs, which facilitates the targeted knowledge transfer from large accurate
teacher models to compact efficient student models. Specifically, the boundary
distillation component focuses on extracting explicit object-level semantic
boundaries from the hierarchical feature maps of the backbone network to
enhance the student model's mask quality in boundary regions. Concurrently, the
context distillation component leverages self-relations as a bridge to transfer
implicit pixel-level contexts from the teacher model to the student model,
ensuring strong connectivity in target regions. Our proposed BCD method is
specifically designed for EDIP tasks and is characterized by its simplicity and
efficiency. Extensive experimental results across semantic segmentation, object
detection, and instance segmentation on various representative datasets
demonstrate that our method can outperform existing methods without requiring
extra supervisions or incurring increased inference costs, resulting in
well-defined object boundaries and smooth connecting regions.
| [
{
"version": "v1",
"created": "Wed, 24 Jan 2024 01:41:26 GMT"
},
{
"version": "v2",
"created": "Mon, 2 Dec 2024 02:55:29 GMT"
},
{
"version": "v3",
"created": "Thu, 27 Mar 2025 01:07:52 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Zhang",
"Dong",
""
],
[
"Dong",
"Pingcheng",
""
],
[
"Chen",
"Long",
""
],
[
"Cheng",
"Kwang-Ting",
""
]
] | TITLE: Towards Complementary Knowledge Distillation for Efficient Dense Image
Prediction
ABSTRACT: It has been revealed that small efficient dense image prediction (EDIP)
models, trained using the knowledge distillation (KD) framework, encounter two
key challenges, including maintaining boundary region completeness and
preserving target region connectivity, despite their favorable capacity to
recognize main object regions. In this work, we propose a complementary
boundary and context distillation (BCD) method within the KD framework for
EDIPs, which facilitates the targeted knowledge transfer from large accurate
teacher models to compact efficient student models. Specifically, the boundary
distillation component focuses on extracting explicit object-level semantic
boundaries from the hierarchical feature maps of the backbone network to
enhance the student model's mask quality in boundary regions. Concurrently, the
context distillation component leverages self-relations as a bridge to transfer
implicit pixel-level contexts from the teacher model to the student model,
ensuring strong connectivity in target regions. Our proposed BCD method is
specifically designed for EDIP tasks and is characterized by its simplicity and
efficiency. Extensive experimental results across semantic segmentation, object
detection, and instance segmentation on various representative datasets
demonstrate that our method can outperform existing methods without requiring
extra supervisions or incurring increased inference costs, resulting in
well-defined object boundaries and smooth connecting regions.
|
2403.12922 | Hanlin Wang | Hanlin Wang, Zhan Tong, Kecheng Zheng, Yujun Shen and Limin Wang | Contextual AD Narration with Interleaved Multimodal Sequence | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Audio Description (AD) task aims to generate descriptions of visual
elements for visually impaired individuals to help them access long-form video
content, like movies. With video feature, text, character bank and context
information as inputs, the generated ADs are able to correspond to the
characters by name and provide reasonable, contextual descriptions to help
audience understand the storyline of movie. To achieve this goal, we propose to
leverage pre-trained foundation models through a simple and unified framework
to generate ADs with interleaved multimodal sequence as input, termed as
Uni-AD. To enhance the alignment of features across various modalities with
finer granularity, we introduce a simple and lightweight module that maps video
features into the textual feature space. Moreover, we also propose a
character-refinement module to provide more precise information by identifying
the main characters who play more significant roles in the video context. With
these unique designs, we further incorporate contextual information and a
contrastive loss into our architecture to generate smoother and more
contextually appropriate ADs. Experiments on multiple AD datasets show that
Uni-AD performs well on AD generation, which demonstrates the effectiveness of
our approach. Our code is available at: https://github.com/ant-research/UniAD.
| [
{
"version": "v1",
"created": "Tue, 19 Mar 2024 17:27:55 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 14:51:25 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Wang",
"Hanlin",
""
],
[
"Tong",
"Zhan",
""
],
[
"Zheng",
"Kecheng",
""
],
[
"Shen",
"Yujun",
""
],
[
"Wang",
"Limin",
""
]
] | TITLE: Contextual AD Narration with Interleaved Multimodal Sequence
ABSTRACT: The Audio Description (AD) task aims to generate descriptions of visual
elements for visually impaired individuals to help them access long-form video
content, like movies. With video feature, text, character bank and context
information as inputs, the generated ADs are able to correspond to the
characters by name and provide reasonable, contextual descriptions to help
audience understand the storyline of movie. To achieve this goal, we propose to
leverage pre-trained foundation models through a simple and unified framework
to generate ADs with interleaved multimodal sequence as input, termed as
Uni-AD. To enhance the alignment of features across various modalities with
finer granularity, we introduce a simple and lightweight module that maps video
features into the textual feature space. Moreover, we also propose a
character-refinement module to provide more precise information by identifying
the main characters who play more significant roles in the video context. With
these unique designs, we further incorporate contextual information and a
contrastive loss into our architecture to generate smoother and more
contextually appropriate ADs. Experiments on multiple AD datasets show that
Uni-AD performs well on AD generation, which demonstrates the effectiveness of
our approach. Our code is available at: https://github.com/ant-research/UniAD.
|
2405.15474 | Gongxi Zhu | Hanlin Gu, Gongxi Zhu, Jie Zhang, Xinyuan Zhao, Yuxing Han, Lixin Fan,
Qiang Yang | Unlearning during Learning: An Efficient Federated Machine Unlearning
Method | Accepted by IJCAI 2024 | null | null | null | cs.LG cs.DC | http://creativecommons.org/licenses/by-nc-nd/4.0/ | In recent years, Federated Learning (FL) has garnered significant attention
as a distributed machine learning paradigm. To facilitate the implementation of
the right to be forgotten, the concept of federated machine unlearning (FMU)
has also emerged. However, current FMU approaches often involve additional
time-consuming steps and may not offer comprehensive unlearning capabilities,
which renders them less practical in real FL scenarios. In this paper, we
introduce FedAU, an innovative and efficient FMU framework aimed at overcoming
these limitations. Specifically, FedAU incorporates a lightweight auxiliary
unlearning module into the learning process and employs a straightforward
linear operation to facilitate unlearning. This approach eliminates the
requirement for extra time-consuming steps, rendering it well-suited for FL.
Furthermore, FedAU exhibits remarkable versatility. It not only enables
multiple clients to carry out unlearning tasks concurrently but also supports
unlearning at various levels of granularity, including individual data samples,
specific classes, and even at the client level. We conducted extensive
experiments on MNIST, CIFAR10, and CIFAR100 datasets to evaluate the
performance of FedAU. The results demonstrate that FedAU effectively achieves
the desired unlearning effect while maintaining model accuracy. Our code is
availiable at https://github.com/Liar-Mask/FedAU.
| [
{
"version": "v1",
"created": "Fri, 24 May 2024 11:53:13 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 12:41:08 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Gu",
"Hanlin",
""
],
[
"Zhu",
"Gongxi",
""
],
[
"Zhang",
"Jie",
""
],
[
"Zhao",
"Xinyuan",
""
],
[
"Han",
"Yuxing",
""
],
[
"Fan",
"Lixin",
""
],
[
"Yang",
"Qiang",
""
]
] | TITLE: Unlearning during Learning: An Efficient Federated Machine Unlearning
Method
ABSTRACT: In recent years, Federated Learning (FL) has garnered significant attention
as a distributed machine learning paradigm. To facilitate the implementation of
the right to be forgotten, the concept of federated machine unlearning (FMU)
has also emerged. However, current FMU approaches often involve additional
time-consuming steps and may not offer comprehensive unlearning capabilities,
which renders them less practical in real FL scenarios. In this paper, we
introduce FedAU, an innovative and efficient FMU framework aimed at overcoming
these limitations. Specifically, FedAU incorporates a lightweight auxiliary
unlearning module into the learning process and employs a straightforward
linear operation to facilitate unlearning. This approach eliminates the
requirement for extra time-consuming steps, rendering it well-suited for FL.
Furthermore, FedAU exhibits remarkable versatility. It not only enables
multiple clients to carry out unlearning tasks concurrently but also supports
unlearning at various levels of granularity, including individual data samples,
specific classes, and even at the client level. We conducted extensive
experiments on MNIST, CIFAR10, and CIFAR100 datasets to evaluate the
performance of FedAU. The results demonstrate that FedAU effectively achieves
the desired unlearning effect while maintaining model accuracy. Our code is
availiable at https://github.com/Liar-Mask/FedAU.
|
2405.15668 | Mahmoud Afifi | Abdelrahman Abdelhamed, Mahmoud Afifi, Alec Go | What Do You See? Enhancing Zero-Shot Image Classification with
Multimodal Large Language Models | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large language models (LLMs) have been effectively used for many computer
vision tasks, including image classification. In this paper, we present a
simple yet effective approach for zero-shot image classification using
multimodal LLMs. Using multimodal LLMs, we generate comprehensive textual
representations from input images. These textual representations are then
utilized to generate fixed-dimensional features in a cross-modal embedding
space. Subsequently, these features are fused together to perform zero-shot
classification using a linear classifier. Our method does not require prompt
engineering for each dataset; instead, we use a single, straightforward set of
prompts across all datasets. We evaluated our method on several datasets and
our results demonstrate its remarkable effectiveness, surpassing benchmark
accuracy on multiple datasets. On average, for ten benchmarks, our method
achieved an accuracy gain of 6.2 percentage points, with an increase of 6.8
percentage points on the ImageNet dataset, compared to prior methods
re-evaluated with the same setup. Our findings highlight the potential of
multimodal LLMs to enhance computer vision tasks such as zero-shot image
classification, offering a significant improvement over traditional methods.
| [
{
"version": "v1",
"created": "Fri, 24 May 2024 16:05:15 GMT"
},
{
"version": "v2",
"created": "Thu, 3 Oct 2024 22:53:09 GMT"
},
{
"version": "v3",
"created": "Sat, 8 Mar 2025 18:53:47 GMT"
},
{
"version": "v4",
"created": "Thu, 27 Mar 2025 09:41:01 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Abdelhamed",
"Abdelrahman",
""
],
[
"Afifi",
"Mahmoud",
""
],
[
"Go",
"Alec",
""
]
] | TITLE: What Do You See? Enhancing Zero-Shot Image Classification with
Multimodal Large Language Models
ABSTRACT: Large language models (LLMs) have been effectively used for many computer
vision tasks, including image classification. In this paper, we present a
simple yet effective approach for zero-shot image classification using
multimodal LLMs. Using multimodal LLMs, we generate comprehensive textual
representations from input images. These textual representations are then
utilized to generate fixed-dimensional features in a cross-modal embedding
space. Subsequently, these features are fused together to perform zero-shot
classification using a linear classifier. Our method does not require prompt
engineering for each dataset; instead, we use a single, straightforward set of
prompts across all datasets. We evaluated our method on several datasets and
our results demonstrate its remarkable effectiveness, surpassing benchmark
accuracy on multiple datasets. On average, for ten benchmarks, our method
achieved an accuracy gain of 6.2 percentage points, with an increase of 6.8
percentage points on the ImageNet dataset, compared to prior methods
re-evaluated with the same setup. Our findings highlight the potential of
multimodal LLMs to enhance computer vision tasks such as zero-shot image
classification, offering a significant improvement over traditional methods.
|
2405.16439 | Rohan Chandra | Rohan Chandra, Haresh Karnan, Negar Mehr, Peter Stone, Joydeep Biswas | Multi-Agent Inverse Reinforcement Learning in Real World Unstructured
Pedestrian Crowds | null | null | null | null | cs.RO cs.AI cs.LG cs.MA | http://creativecommons.org/licenses/by/4.0/ | Social robot navigation in crowded public spaces such as university campuses,
restaurants, grocery stores, and hospitals, is an increasingly important area
of research. One of the core strategies for achieving this goal is to
understand humans' intent--underlying psychological factors that govern their
motion--by learning their reward functions, typically via inverse reinforcement
learning (IRL). Despite significant progress in IRL, learning reward functions
of multiple agents simultaneously in dense unstructured pedestrian crowds has
remained intractable due to the nature of the tightly coupled social
interactions that occur in these scenarios \textit{e.g.} passing,
intersections, swerving, weaving, etc. In this paper, we present a new
multi-agent maximum entropy inverse reinforcement learning algorithm for real
world unstructured pedestrian crowds. Key to our approach is a simple, but
effective, mathematical trick which we name the so-called
tractability-rationality trade-off trick that achieves tractability at the cost
of a slight reduction in accuracy. We compare our approach to the classical
single-agent MaxEnt IRL as well as state-of-the-art trajectory prediction
methods on several datasets including the ETH, UCY, SCAND, JRDB, and a new
dataset, called Speedway, collected at a busy intersection on a University
campus focusing on dense, complex agent interactions. Our key findings show
that, on the dense Speedway dataset, our approach ranks 1st among top 7
baselines with >2X improvement over single-agent IRL, and is competitive with
state-of-the-art large transformer-based encoder-decoder models on sparser
datasets such as ETH/UCY (ranks 3rd among top 7 baselines).
| [
{
"version": "v1",
"created": "Sun, 26 May 2024 05:48:21 GMT"
},
{
"version": "v2",
"created": "Sun, 15 Dec 2024 03:48:49 GMT"
},
{
"version": "v3",
"created": "Wed, 26 Mar 2025 21:19:58 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Chandra",
"Rohan",
""
],
[
"Karnan",
"Haresh",
""
],
[
"Mehr",
"Negar",
""
],
[
"Stone",
"Peter",
""
],
[
"Biswas",
"Joydeep",
""
]
] | TITLE: Multi-Agent Inverse Reinforcement Learning in Real World Unstructured
Pedestrian Crowds
ABSTRACT: Social robot navigation in crowded public spaces such as university campuses,
restaurants, grocery stores, and hospitals, is an increasingly important area
of research. One of the core strategies for achieving this goal is to
understand humans' intent--underlying psychological factors that govern their
motion--by learning their reward functions, typically via inverse reinforcement
learning (IRL). Despite significant progress in IRL, learning reward functions
of multiple agents simultaneously in dense unstructured pedestrian crowds has
remained intractable due to the nature of the tightly coupled social
interactions that occur in these scenarios \textit{e.g.} passing,
intersections, swerving, weaving, etc. In this paper, we present a new
multi-agent maximum entropy inverse reinforcement learning algorithm for real
world unstructured pedestrian crowds. Key to our approach is a simple, but
effective, mathematical trick which we name the so-called
tractability-rationality trade-off trick that achieves tractability at the cost
of a slight reduction in accuracy. We compare our approach to the classical
single-agent MaxEnt IRL as well as state-of-the-art trajectory prediction
methods on several datasets including the ETH, UCY, SCAND, JRDB, and a new
dataset, called Speedway, collected at a busy intersection on a University
campus focusing on dense, complex agent interactions. Our key findings show
that, on the dense Speedway dataset, our approach ranks 1st among top 7
baselines with >2X improvement over single-agent IRL, and is competitive with
state-of-the-art large transformer-based encoder-decoder models on sparser
datasets such as ETH/UCY (ranks 3rd among top 7 baselines).
|
2405.17712 | Mohammad Hasan Dr. | Ahatsham Hayat and Mohammad Rashedul Hasan | A Context-Aware Approach for Enhancing Data Imputation with Pre-trained
Language Models | null | null | null | null | cs.CL cs.LG | http://creativecommons.org/licenses/by/4.0/ | This paper presents a novel approach named \textbf{C}ontextually
\textbf{R}elevant \textbf{I}mputation leveraging pre-trained \textbf{L}anguage
\textbf{M}odels (\textbf{CRILM}) for handling missing data in tabular datasets.
Instead of relying on traditional numerical estimations, CRILM uses pre-trained
language models (LMs) to create contextually relevant descriptors for missing
values. This method aligns datasets with LMs' strengths, allowing large LMs to
generate these descriptors and small LMs to be fine-tuned on the enriched
datasets for enhanced downstream task performance. Our evaluations demonstrate
CRILM's superior performance and robustness across MCAR, MAR, and challenging
MNAR scenarios, with up to a 10\% improvement over the best-performing
baselines. By mitigating biases, particularly in MNAR settings, CRILM improves
downstream task performance and offers a cost-effective solution for
resource-constrained environments.
| [
{
"version": "v1",
"created": "Tue, 28 May 2024 00:08:29 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 16:22:43 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Hayat",
"Ahatsham",
""
],
[
"Hasan",
"Mohammad Rashedul",
""
]
] | TITLE: A Context-Aware Approach for Enhancing Data Imputation with Pre-trained
Language Models
ABSTRACT: This paper presents a novel approach named \textbf{C}ontextually
\textbf{R}elevant \textbf{I}mputation leveraging pre-trained \textbf{L}anguage
\textbf{M}odels (\textbf{CRILM}) for handling missing data in tabular datasets.
Instead of relying on traditional numerical estimations, CRILM uses pre-trained
language models (LMs) to create contextually relevant descriptors for missing
values. This method aligns datasets with LMs' strengths, allowing large LMs to
generate these descriptors and small LMs to be fine-tuned on the enriched
datasets for enhanced downstream task performance. Our evaluations demonstrate
CRILM's superior performance and robustness across MCAR, MAR, and challenging
MNAR scenarios, with up to a 10\% improvement over the best-performing
baselines. By mitigating biases, particularly in MNAR settings, CRILM improves
downstream task performance and offers a cost-effective solution for
resource-constrained environments.
|
2406.02166 | Saierdaer Yusuyin | Saierdaer Yusuyin, Te Ma, Hao Huang, Wenbo Zhao, Zhijian Ou | Whistle: Data-Efficient Multilingual and Crosslingual Speech Recognition
via Weakly Phonetic Supervision | Accepted by IEEE-TASLP | null | 10.1109/TASLPRO.2025.3550683 | null | cs.SD cs.CL eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | There exist three approaches for multilingual and crosslingual automatic
speech recognition (MCL-ASR) - supervised pretraining with phonetic or
graphemic transcription, and self-supervised pretraining. We find that
pretraining with phonetic supervision has been underappreciated so far for
MCL-ASR, while conceptually it is more advantageous for information sharing
between different languages. This paper explores the approach of pretraining
with weakly phonetic supervision towards data-efficient MCL-ASR, which is
called Whistle. We relax the requirement of gold-standard human-validated
phonetic transcripts, and obtain International Phonetic Alphabet (IPA) based
transcription by leveraging the LanguageNet grapheme-to-phoneme (G2P) models.
We construct a common experimental setup based on the CommonVoice dataset,
called CV-Lang10, with 10 seen languages and 2 unseen languages. A set of
experiments are conducted on CV-Lang10 to compare, as fair as possible, the
three approaches under the common setup for MCL-ASR. Experiments demonstrate
the advantages of phoneme-based models (Whistle) for MCL-ASR, in terms of
speech recognition for seen languages, crosslingual performance for unseen
languages with different amounts of few-shot data, overcoming catastrophic
forgetting, and training efficiency. It is found that when training data is
more limited, phoneme supervision can achieve better results compared to
subword supervision and self-supervision, thereby providing higher
data-efficiency. To support reproducibility and promote future research along
this direction, we release the code, models and data for the entire pipeline of
Whistle at https://github.com/thu-spmi/CAT/tree/master/egs/cv-lang10.
| [
{
"version": "v1",
"created": "Tue, 4 Jun 2024 09:56:05 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 16:38:29 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Yusuyin",
"Saierdaer",
""
],
[
"Ma",
"Te",
""
],
[
"Huang",
"Hao",
""
],
[
"Zhao",
"Wenbo",
""
],
[
"Ou",
"Zhijian",
""
]
] | TITLE: Whistle: Data-Efficient Multilingual and Crosslingual Speech Recognition
via Weakly Phonetic Supervision
ABSTRACT: There exist three approaches for multilingual and crosslingual automatic
speech recognition (MCL-ASR) - supervised pretraining with phonetic or
graphemic transcription, and self-supervised pretraining. We find that
pretraining with phonetic supervision has been underappreciated so far for
MCL-ASR, while conceptually it is more advantageous for information sharing
between different languages. This paper explores the approach of pretraining
with weakly phonetic supervision towards data-efficient MCL-ASR, which is
called Whistle. We relax the requirement of gold-standard human-validated
phonetic transcripts, and obtain International Phonetic Alphabet (IPA) based
transcription by leveraging the LanguageNet grapheme-to-phoneme (G2P) models.
We construct a common experimental setup based on the CommonVoice dataset,
called CV-Lang10, with 10 seen languages and 2 unseen languages. A set of
experiments are conducted on CV-Lang10 to compare, as fair as possible, the
three approaches under the common setup for MCL-ASR. Experiments demonstrate
the advantages of phoneme-based models (Whistle) for MCL-ASR, in terms of
speech recognition for seen languages, crosslingual performance for unseen
languages with different amounts of few-shot data, overcoming catastrophic
forgetting, and training efficiency. It is found that when training data is
more limited, phoneme supervision can achieve better results compared to
subword supervision and self-supervision, thereby providing higher
data-efficiency. To support reproducibility and promote future research along
this direction, we release the code, models and data for the entire pipeline of
Whistle at https://github.com/thu-spmi/CAT/tree/master/egs/cv-lang10.
|
2407.03314 | Zhantao Yang | Zhantao Yang, Ruili Feng, Keyu Yan, Huangji Wang, Zhicai Wang,
Shangwen Zhu, Han Zhang, Jie Xiao, Pingyu Wu, Kai Zhu, Jixuan Chen, Chen-Wei
Xie, Yue Yang, Hongyang Zhang, Yu Liu, Fan Cheng | BACON: Improving Clarity of Image Captions via Bag-of-Concept Graphs | null | null | null | null | cs.CV cs.CL cs.DB | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Advancements in large Vision-Language Models have brought precise, accurate
image captioning, vital for advancing multi-modal image understanding and
processing. Yet these captions often carry lengthy, intertwined contexts that
are difficult to parse and frequently overlook essential cues, posing a great
barrier for models like GroundingDINO and SDXL, which lack the strong text
encoding and syntax analysis needed to fully leverage dense captions. To
address this, we propose BACON, a prompting method that breaks down
VLM-generated captions into disentangled, structured elements such as objects,
relationships, styles, and themes. This approach not only minimizes confusion
from handling complex contexts but also allows for efficient transfer into a
JSON dictionary, enabling models without linguistic processing capabilities to
easily access key information. We annotated 100,000 image-caption pairs using
BACON with GPT-4V and trained an LLaVA captioner on this dataset, enabling it
to produce BACON-style captions without relying on costly GPT-4V. Evaluations
of overall quality, precision, and recall-as well as user studies-demonstrate
that the resulting caption model consistently outperforms other SOTA VLM models
in generating high-quality captions. Besides, we show that BACON-style captions
exhibit better clarity when applied to various models, enabling them to
accomplish previously unattainable tasks or surpass existing SOTA solutions
without training. For example, BACON-style captions help GroundingDINO achieve
1.51x higher recall scores on open-vocabulary object detection tasks compared
to leading methods.
| [
{
"version": "v1",
"created": "Wed, 3 Jul 2024 17:55:27 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 17:06:25 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Yang",
"Zhantao",
""
],
[
"Feng",
"Ruili",
""
],
[
"Yan",
"Keyu",
""
],
[
"Wang",
"Huangji",
""
],
[
"Wang",
"Zhicai",
""
],
[
"Zhu",
"Shangwen",
""
],
[
"Zhang",
"Han",
""
],
[
"Xiao",
"Jie",
""
],
[
"Wu",
"Pingyu",
""
],
[
"Zhu",
"Kai",
""
],
[
"Chen",
"Jixuan",
""
],
[
"Xie",
"Chen-Wei",
""
],
[
"Yang",
"Yue",
""
],
[
"Zhang",
"Hongyang",
""
],
[
"Liu",
"Yu",
""
],
[
"Cheng",
"Fan",
""
]
] | TITLE: BACON: Improving Clarity of Image Captions via Bag-of-Concept Graphs
ABSTRACT: Advancements in large Vision-Language Models have brought precise, accurate
image captioning, vital for advancing multi-modal image understanding and
processing. Yet these captions often carry lengthy, intertwined contexts that
are difficult to parse and frequently overlook essential cues, posing a great
barrier for models like GroundingDINO and SDXL, which lack the strong text
encoding and syntax analysis needed to fully leverage dense captions. To
address this, we propose BACON, a prompting method that breaks down
VLM-generated captions into disentangled, structured elements such as objects,
relationships, styles, and themes. This approach not only minimizes confusion
from handling complex contexts but also allows for efficient transfer into a
JSON dictionary, enabling models without linguistic processing capabilities to
easily access key information. We annotated 100,000 image-caption pairs using
BACON with GPT-4V and trained an LLaVA captioner on this dataset, enabling it
to produce BACON-style captions without relying on costly GPT-4V. Evaluations
of overall quality, precision, and recall-as well as user studies-demonstrate
that the resulting caption model consistently outperforms other SOTA VLM models
in generating high-quality captions. Besides, we show that BACON-style captions
exhibit better clarity when applied to various models, enabling them to
accomplish previously unattainable tasks or surpass existing SOTA solutions
without training. For example, BACON-style captions help GroundingDINO achieve
1.51x higher recall scores on open-vocabulary object detection tasks compared
to leading methods.
|
2407.05608 | Xiaoxiao Miao | Xiaoxiao Miao, Ruijie Tao, Chang Zeng, Xin Wang | A Benchmark for Multi-speaker Anonymization | Accepted by TIFS | null | null | null | cs.SD cs.CL eess.AS | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Privacy-preserving voice protection approaches primarily suppress
privacy-related information derived from paralinguistic attributes while
preserving the linguistic content. Existing solutions focus particularly on
single-speaker scenarios. However, they lack practicality for real-world
applications, i.e., multi-speaker scenarios. In this paper, we present an
initial attempt to provide a multi-speaker anonymization benchmark by defining
the task and evaluation protocol, proposing benchmarking solutions, and
discussing the privacy leakage of overlapping conversations. The proposed
benchmark solutions are based on a cascaded system that integrates
spectral-clustering-based speaker diarization and disentanglement-based speaker
anonymization using a selection-based anonymizer. To improve utility, the
benchmark solutions are further enhanced by two conversation-level speaker
vector anonymization methods. The first method minimizes the differential
similarity across speaker pairs in the original and anonymized conversations,
which maintains original speaker relationships in the anonymized version. The
other minimizes the aggregated similarity across anonymized speakers, which
achieves better differentiation between speakers.Experiments conducted on both
non-overlap simulated and real-world datasets demonstrate the effectiveness of
the multi-speaker anonymization system with the proposed speaker anonymizers.
Additionally, we analyzed overlapping speech regarding privacy leakage and
provided potential solutions
| [
{
"version": "v1",
"created": "Mon, 8 Jul 2024 04:48:43 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 06:27:57 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Miao",
"Xiaoxiao",
""
],
[
"Tao",
"Ruijie",
""
],
[
"Zeng",
"Chang",
""
],
[
"Wang",
"Xin",
""
]
] | TITLE: A Benchmark for Multi-speaker Anonymization
ABSTRACT: Privacy-preserving voice protection approaches primarily suppress
privacy-related information derived from paralinguistic attributes while
preserving the linguistic content. Existing solutions focus particularly on
single-speaker scenarios. However, they lack practicality for real-world
applications, i.e., multi-speaker scenarios. In this paper, we present an
initial attempt to provide a multi-speaker anonymization benchmark by defining
the task and evaluation protocol, proposing benchmarking solutions, and
discussing the privacy leakage of overlapping conversations. The proposed
benchmark solutions are based on a cascaded system that integrates
spectral-clustering-based speaker diarization and disentanglement-based speaker
anonymization using a selection-based anonymizer. To improve utility, the
benchmark solutions are further enhanced by two conversation-level speaker
vector anonymization methods. The first method minimizes the differential
similarity across speaker pairs in the original and anonymized conversations,
which maintains original speaker relationships in the anonymized version. The
other minimizes the aggregated similarity across anonymized speakers, which
achieves better differentiation between speakers.Experiments conducted on both
non-overlap simulated and real-world datasets demonstrate the effectiveness of
the multi-speaker anonymization system with the proposed speaker anonymizers.
Additionally, we analyzed overlapping speech regarding privacy leakage and
provided potential solutions
|
2407.11828 | Julien Hauret | Julien Hauret and Malo Olivier and Thomas Joubaud and Christophe
Langrenne and Sarah Poir\'ee and V\'eronique Zimpfer and \'Eric Bavu | Vibravox: A Dataset of French Speech Captured with Body-conduction Audio
Sensors | 23 pages, 42 figures | null | null | null | eess.AS cs.LG | http://creativecommons.org/licenses/by/4.0/ | Vibravox is a dataset compliant with the General Data Protection Regulation
(GDPR) containing audio recordings using five different body-conduction audio
sensors: two in-ear microphones, two bone conduction vibration pickups, and a
laryngophone. The dataset also includes audio data from an airborne microphone
used as a reference. The Vibravox corpus contains 45 hours per sensor of speech
samples and physiological sounds recorded by 188 participants under different
acoustic conditions imposed by a high order ambisonics 3D spatializer.
Annotations about the recording conditions and linguistic transcriptions are
also included in the corpus. We conducted a series of experiments on various
speech-related tasks, including speech recognition, speech enhancement, and
speaker verification. These experiments were carried out using state-of-the-art
models to evaluate and compare their performances on signals captured by the
different audio sensors offered by the Vibravox dataset, with the aim of
gaining a better grasp of their individual characteristics.
| [
{
"version": "v1",
"created": "Tue, 16 Jul 2024 15:16:10 GMT"
},
{
"version": "v2",
"created": "Wed, 17 Jul 2024 08:09:01 GMT"
},
{
"version": "v3",
"created": "Fri, 21 Feb 2025 17:42:56 GMT"
},
{
"version": "v4",
"created": "Thu, 27 Mar 2025 01:13:48 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Hauret",
"Julien",
""
],
[
"Olivier",
"Malo",
""
],
[
"Joubaud",
"Thomas",
""
],
[
"Langrenne",
"Christophe",
""
],
[
"Poirée",
"Sarah",
""
],
[
"Zimpfer",
"Véronique",
""
],
[
"Bavu",
"Éric",
""
]
] | TITLE: Vibravox: A Dataset of French Speech Captured with Body-conduction Audio
Sensors
ABSTRACT: Vibravox is a dataset compliant with the General Data Protection Regulation
(GDPR) containing audio recordings using five different body-conduction audio
sensors: two in-ear microphones, two bone conduction vibration pickups, and a
laryngophone. The dataset also includes audio data from an airborne microphone
used as a reference. The Vibravox corpus contains 45 hours per sensor of speech
samples and physiological sounds recorded by 188 participants under different
acoustic conditions imposed by a high order ambisonics 3D spatializer.
Annotations about the recording conditions and linguistic transcriptions are
also included in the corpus. We conducted a series of experiments on various
speech-related tasks, including speech recognition, speech enhancement, and
speaker verification. These experiments were carried out using state-of-the-art
models to evaluate and compare their performances on signals captured by the
different audio sensors offered by the Vibravox dataset, with the aim of
gaining a better grasp of their individual characteristics.
|
2408.00279 | Yesheng Zhang | Yesheng Zhang, Shuhan Shen, Xu Zhao | MESA: Effective Matching Redundancy Reduction by Semantic Area
Segmentation | 18pages+suppl | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | We propose MESA and DMESA as novel feature matching methods, which utilize
Segment Anything Model (SAM) to effectively mitigate matching redundancy. The
key insight of our methods is to establish implicit-semantic area matching
prior to point matching, based on advanced image understanding of SAM. Then,
informative area matches with consistent internal semantic are able to undergo
dense feature comparison, facilitating precise inside-area point matching.
Specifically, MESA adopts a sparse matching framework and first obtains
candidate areas from SAM results through a novel Area Graph (AG). Then, area
matching among the candidates is formulated as graph energy minimization and
solved by graphical models derived from AG. To address the efficiency issue of
MESA, we further propose DMESA as its dense counterpart, applying a dense
matching framework. After candidate areas are identified by AG, DMESA
establishes area matches through generating dense matching distributions. The
distributions are produced from off-the-shelf patch matching utilizing the
Gaussian Mixture Model and refined via the Expectation Maximization. With less
repetitive computation, DMESA showcases a speed improvement of nearly five
times compared to MESA, while maintaining competitive accuracy. Our methods are
extensively evaluated on five datasets encompassing indoor and outdoor scenes.
The results illustrate consistent performance improvements from our methods for
five distinct point matching baselines across all datasets. Furthermore, our
methods exhibit promise generalization and improved robustness against image
resolution variations. The code is publicly available at
https://github.com/Easonyesheng/A2PM-MESA.
| [
{
"version": "v1",
"created": "Thu, 1 Aug 2024 04:39:36 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 08:11:57 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Zhang",
"Yesheng",
""
],
[
"Shen",
"Shuhan",
""
],
[
"Zhao",
"Xu",
""
]
] | TITLE: MESA: Effective Matching Redundancy Reduction by Semantic Area
Segmentation
ABSTRACT: We propose MESA and DMESA as novel feature matching methods, which utilize
Segment Anything Model (SAM) to effectively mitigate matching redundancy. The
key insight of our methods is to establish implicit-semantic area matching
prior to point matching, based on advanced image understanding of SAM. Then,
informative area matches with consistent internal semantic are able to undergo
dense feature comparison, facilitating precise inside-area point matching.
Specifically, MESA adopts a sparse matching framework and first obtains
candidate areas from SAM results through a novel Area Graph (AG). Then, area
matching among the candidates is formulated as graph energy minimization and
solved by graphical models derived from AG. To address the efficiency issue of
MESA, we further propose DMESA as its dense counterpart, applying a dense
matching framework. After candidate areas are identified by AG, DMESA
establishes area matches through generating dense matching distributions. The
distributions are produced from off-the-shelf patch matching utilizing the
Gaussian Mixture Model and refined via the Expectation Maximization. With less
repetitive computation, DMESA showcases a speed improvement of nearly five
times compared to MESA, while maintaining competitive accuracy. Our methods are
extensively evaluated on five datasets encompassing indoor and outdoor scenes.
The results illustrate consistent performance improvements from our methods for
five distinct point matching baselines across all datasets. Furthermore, our
methods exhibit promise generalization and improved robustness against image
resolution variations. The code is publicly available at
https://github.com/Easonyesheng/A2PM-MESA.
|
2408.09769 | Enrico del Re | Enrico Del Re, Amirhesam Aghanouri, Cristina Olaverri-Monreal | Integrating Naturalistic Insights in Objective Multi-Vehicle Safety
Framework | null | null | 10.1109/ITSC58415.2024.10920258. | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | As autonomous vehicle technology advances, the precise assessment of safety
in complex traffic scenarios becomes crucial, especially in mixed-vehicle
environments where human perception of safety must be taken into account. This
paper presents a framework designed for assessing traffic safety in
multi-vehicle situations, facilitating the simultaneous utilization of diverse
objective safety metrics. Additionally, it allows the integration of subjective
perception of safety by adjusting model parameters. The framework was applied
to evaluate various model configurations in car-following scenarios on a
highway, utilizing naturalistic driving datasets. The evaluation of the model
showed an outstanding performance, particularly when integrating multiple
objective safety measures. Furthermore, the performance was significantly
enhanced when considering all surrounding vehicles.
| [
{
"version": "v1",
"created": "Mon, 19 Aug 2024 07:58:10 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 12:09:05 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Del Re",
"Enrico",
""
],
[
"Aghanouri",
"Amirhesam",
""
],
[
"Olaverri-Monreal",
"Cristina",
""
]
] | TITLE: Integrating Naturalistic Insights in Objective Multi-Vehicle Safety
Framework
ABSTRACT: As autonomous vehicle technology advances, the precise assessment of safety
in complex traffic scenarios becomes crucial, especially in mixed-vehicle
environments where human perception of safety must be taken into account. This
paper presents a framework designed for assessing traffic safety in
multi-vehicle situations, facilitating the simultaneous utilization of diverse
objective safety metrics. Additionally, it allows the integration of subjective
perception of safety by adjusting model parameters. The framework was applied
to evaluate various model configurations in car-following scenarios on a
highway, utilizing naturalistic driving datasets. The evaluation of the model
showed an outstanding performance, particularly when integrating multiple
objective safety measures. Furthermore, the performance was significantly
enhanced when considering all surrounding vehicles.
|
2408.09833 | Mohamed Sabry MSc | Mohamed Sabry, Walter Morales-Alvarez and Cristina Olaverri-Monreal | Automated Vehicle Driver Monitoring Dataset from Real-World Scenarios | 6 pages | null | 10.1109/ITSC58415.2024.10920048 | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | From SAE Level 3 of automation onwards, drivers are allowed to engage in
activities that are not directly related to driving during their travel.
However, in level 3, a misunderstanding of the capabilities of the system might
lead drivers to engage in secondary tasks, which could impair their ability to
react to challenging traffic situations.
Anticipating driver activity allows for early detection of risky behaviors,
to prevent accidents. To be able to predict the driver activity, a Deep
Learning network needs to be trained on a dataset. However, the use of datasets
based on simulation for training and the migration to real-world data for
prediction has proven to be suboptimal. Hence, this paper presents a real-world
driver activity dataset, openly accessible on IEEE Dataport, which encompasses
various activities that occur in autonomous driving scenarios under various
illumination and weather conditions. Results from the training process showed
that the dataset provides an excellent benchmark for implementing models for
driver activity recognition.
| [
{
"version": "v1",
"created": "Mon, 19 Aug 2024 09:29:00 GMT"
},
{
"version": "v2",
"created": "Wed, 26 Mar 2025 22:41:51 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Sabry",
"Mohamed",
""
],
[
"Morales-Alvarez",
"Walter",
""
],
[
"Olaverri-Monreal",
"Cristina",
""
]
] | TITLE: Automated Vehicle Driver Monitoring Dataset from Real-World Scenarios
ABSTRACT: From SAE Level 3 of automation onwards, drivers are allowed to engage in
activities that are not directly related to driving during their travel.
However, in level 3, a misunderstanding of the capabilities of the system might
lead drivers to engage in secondary tasks, which could impair their ability to
react to challenging traffic situations.
Anticipating driver activity allows for early detection of risky behaviors,
to prevent accidents. To be able to predict the driver activity, a Deep
Learning network needs to be trained on a dataset. However, the use of datasets
based on simulation for training and the migration to real-world data for
prediction has proven to be suboptimal. Hence, this paper presents a real-world
driver activity dataset, openly accessible on IEEE Dataport, which encompasses
various activities that occur in autonomous driving scenarios under various
illumination and weather conditions. Results from the training process showed
that the dataset provides an excellent benchmark for implementing models for
driver activity recognition.
|
2408.12691 | Pooya Ashtari | Pooya Ashtari, Pourya Behmandpoor, Fateme Nateghi Haredasht, Jonathan
H. Chen, Panagiotis Patrinos and Sabine Van Huffel | Quantization-aware Matrix Factorization for Low Bit Rate Image
Compression | 22 pages, 6 figures, 1 table, 1 algorithm | null | null | null | eess.IV cs.CV math.OC | http://creativecommons.org/licenses/by/4.0/ | Lossy image compression is essential for efficient transmission and storage.
Traditional compression methods mainly rely on discrete cosine transform (DCT)
or singular value decomposition (SVD), both of which represent image data in
continuous domains and, therefore, necessitate carefully designed quantizers.
Notably, these methods consider quantization as a separate step, where
quantization errors cannot be incorporated into the compression process. The
sensitivity of these methods, especially SVD-based ones, to quantization errors
significantly degrades reconstruction quality. To address this issue, we
introduce a quantization-aware matrix factorization (QMF) to develop a novel
lossy image compression method. QMF provides a low-rank representation of the
image data as a product of two smaller factor matrices, with elements
constrained to bounded integer values, thereby effectively integrating
quantization with low-rank approximation. We propose an efficient, provably
convergent iterative algorithm for QMF using a block coordinate descent (BCD)
scheme, with subproblems having closed-form solutions. Our experiments on the
Kodak and CLIC 2024 datasets demonstrate that our QMF compression method
consistently outperforms JPEG at low bit rates below 0.25 bits per pixel (bpp)
and remains comparable at higher bit rates. We also assessed our method's
capability to preserve visual semantics by evaluating an ImageNet pre-trained
classifier on compressed images. Remarkably, our method improved top-1 accuracy
by over 5 percentage points compared to JPEG at bit rates under 0.25 bpp. The
project is available at https://github.com/pashtari/lrf .
| [
{
"version": "v1",
"created": "Thu, 22 Aug 2024 19:08:08 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 14:26:49 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Ashtari",
"Pooya",
""
],
[
"Behmandpoor",
"Pourya",
""
],
[
"Haredasht",
"Fateme Nateghi",
""
],
[
"Chen",
"Jonathan H.",
""
],
[
"Patrinos",
"Panagiotis",
""
],
[
"Van Huffel",
"Sabine",
""
]
] | TITLE: Quantization-aware Matrix Factorization for Low Bit Rate Image
Compression
ABSTRACT: Lossy image compression is essential for efficient transmission and storage.
Traditional compression methods mainly rely on discrete cosine transform (DCT)
or singular value decomposition (SVD), both of which represent image data in
continuous domains and, therefore, necessitate carefully designed quantizers.
Notably, these methods consider quantization as a separate step, where
quantization errors cannot be incorporated into the compression process. The
sensitivity of these methods, especially SVD-based ones, to quantization errors
significantly degrades reconstruction quality. To address this issue, we
introduce a quantization-aware matrix factorization (QMF) to develop a novel
lossy image compression method. QMF provides a low-rank representation of the
image data as a product of two smaller factor matrices, with elements
constrained to bounded integer values, thereby effectively integrating
quantization with low-rank approximation. We propose an efficient, provably
convergent iterative algorithm for QMF using a block coordinate descent (BCD)
scheme, with subproblems having closed-form solutions. Our experiments on the
Kodak and CLIC 2024 datasets demonstrate that our QMF compression method
consistently outperforms JPEG at low bit rates below 0.25 bits per pixel (bpp)
and remains comparable at higher bit rates. We also assessed our method's
capability to preserve visual semantics by evaluating an ImageNet pre-trained
classifier on compressed images. Remarkably, our method improved top-1 accuracy
by over 5 percentage points compared to JPEG at bit rates under 0.25 bpp. The
project is available at https://github.com/pashtari/lrf .
|
2408.16863 | Robert Mahari | Alexandre Mojon, Robert Mahari, Sandro Claudio Lera | Data-Driven Law Firm Rankings to Reduce Information Asymmetry in Legal
Disputes | null | null | null | null | cs.CY | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Selecting capable counsel can shape the outcome of litigation, yet evaluating
law firm performance remains challenging. Widely used rankings prioritize
prestige, size, and revenue rather than empirical litigation outcomes, offering
little practical guidance. To address this gap, we build on the Bradley-Terry
model and introduce a new ranking framework that treats each lawsuit as a
competitive game between plaintiff and defendant law firms. Leveraging a newly
constructed dataset of 60,540 U.S. civil lawsuits involving 54,541 law firms,
our findings show that existing reputation-based rankings correlate poorly with
actual litigation success, whereas our outcome-based ranking substantially
improves predictive accuracy. These findings establish a foundation for more
transparent, data-driven assessments of legal performance.
| [
{
"version": "v1",
"created": "Thu, 29 Aug 2024 19:04:45 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 00:35:30 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Mojon",
"Alexandre",
""
],
[
"Mahari",
"Robert",
""
],
[
"Lera",
"Sandro Claudio",
""
]
] | TITLE: Data-Driven Law Firm Rankings to Reduce Information Asymmetry in Legal
Disputes
ABSTRACT: Selecting capable counsel can shape the outcome of litigation, yet evaluating
law firm performance remains challenging. Widely used rankings prioritize
prestige, size, and revenue rather than empirical litigation outcomes, offering
little practical guidance. To address this gap, we build on the Bradley-Terry
model and introduce a new ranking framework that treats each lawsuit as a
competitive game between plaintiff and defendant law firms. Leveraging a newly
constructed dataset of 60,540 U.S. civil lawsuits involving 54,541 law firms,
our findings show that existing reputation-based rankings correlate poorly with
actual litigation success, whereas our outcome-based ranking substantially
improves predictive accuracy. These findings establish a foundation for more
transparent, data-driven assessments of legal performance.
|
2408.17258 | Tong Nie | Tong Nie, Junlin He, Yuewen Mei, Guoyang Qin, Guilong Li, Jian Sun,
Wei Ma | Joint Estimation and Prediction of City-wide Delivery Demand: A Large
Language Model Empowered Graph-based Learning Approach | null | Transportation Research Part E: Logistics and Transportation
Review, 2025 | 10.1016/j.tre.2025.104075 | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The proliferation of e-commerce and urbanization has significantly
intensified delivery operations in urban areas, boosting the volume and
complexity of delivery demand. Data-driven predictive methods, especially those
utilizing machine learning techniques, have emerged to handle these
complexities in urban delivery demand management problems. One particularly
pressing issue that has yet to be sufficiently addressed is the joint
estimation and prediction of city-wide delivery demand, as well as the
generalization of the model to new cities. To this end, we formulate this
problem as a transferable graph-based spatiotemporal learning task. First, an
individual-collective message-passing neural network model is formalized to
capture the interaction between demand patterns of associated regions. Second,
by exploiting recent advances in large language models (LLMs), we extract
general geospatial knowledge encodings from the unstructured locational data
using the embedding generated by LLMs. Last, to encourage the cross-city
generalization of the model, we integrate the encoding into the demand
predictor in a transferable way. Comprehensive empirical evaluation results on
two real-world delivery datasets, including eight cities in China and the US,
demonstrate that our model significantly outperforms state-of-the-art baselines
in accuracy, efficiency, and transferability.
| [
{
"version": "v1",
"created": "Fri, 30 Aug 2024 12:56:17 GMT"
},
{
"version": "v2",
"created": "Sat, 30 Nov 2024 12:13:01 GMT"
},
{
"version": "v3",
"created": "Thu, 27 Mar 2025 11:41:54 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Nie",
"Tong",
""
],
[
"He",
"Junlin",
""
],
[
"Mei",
"Yuewen",
""
],
[
"Qin",
"Guoyang",
""
],
[
"Li",
"Guilong",
""
],
[
"Sun",
"Jian",
""
],
[
"Ma",
"Wei",
""
]
] | TITLE: Joint Estimation and Prediction of City-wide Delivery Demand: A Large
Language Model Empowered Graph-based Learning Approach
ABSTRACT: The proliferation of e-commerce and urbanization has significantly
intensified delivery operations in urban areas, boosting the volume and
complexity of delivery demand. Data-driven predictive methods, especially those
utilizing machine learning techniques, have emerged to handle these
complexities in urban delivery demand management problems. One particularly
pressing issue that has yet to be sufficiently addressed is the joint
estimation and prediction of city-wide delivery demand, as well as the
generalization of the model to new cities. To this end, we formulate this
problem as a transferable graph-based spatiotemporal learning task. First, an
individual-collective message-passing neural network model is formalized to
capture the interaction between demand patterns of associated regions. Second,
by exploiting recent advances in large language models (LLMs), we extract
general geospatial knowledge encodings from the unstructured locational data
using the embedding generated by LLMs. Last, to encourage the cross-city
generalization of the model, we integrate the encoding into the demand
predictor in a transferable way. Comprehensive empirical evaluation results on
two real-world delivery datasets, including eight cities in China and the US,
demonstrate that our model significantly outperforms state-of-the-art baselines
in accuracy, efficiency, and transferability.
|
2409.09430 | Amirreza Mahbod | Amirreza Mahbod, Nematollah Saeidi, Sepideh Hatamikia, Ramona Woitek | Evaluating Pre-trained Convolutional Neural Networks and Foundation
Models as Feature Extractors for Content-based Medical Image Retrieval | 37 pages | null | 10.1016/j.engappai.2025.110571 | null | cs.CV | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Medical image retrieval refers to the task of finding similar images for
given query images in a database, with applications such as diagnosis support.
While traditional medical image retrieval relied on clinical metadata,
content-based medical image retrieval (CBMIR) depends on image features, which
can be extracted automatically or semi-automatically. Many approaches have been
proposed for CBMIR, and among them, using pre-trained convolutional neural
networks (CNNs) is a widely utilized approach. However, considering the recent
advances in the development of foundation models for various computer vision
tasks, their application for CBMIR can also be investigated.
In this study, we used several pre-trained feature extractors from well-known
pre-trained CNNs and pre-trained foundation models and investigated the CBMIR
performance on eight types of two-dimensional (2D) and three-dimensional (3D)
medical images. Furthermore, we investigated the effect of image size on the
CBMIR performance.
Our results show that, overall, for the 2D datasets, foundation models
deliver superior performance by a large margin compared to CNNs, with the
general-purpose self-supervised model for computational pathology (UNI)
providing the best overall performance across all datasets and image sizes. For
3D datasets, CNNs and foundation models deliver more competitive performance,
with contrastive learning from captions for histopathology model (CONCH)
achieving the best overall performance. Moreover, our findings confirm that
while using larger image sizes (especially for 2D datasets) yields slightly
better performance, competitive CBMIR performance can still be achieved even
with smaller image sizes. Our codes to reproduce the results are available at:
https://github.com/masih4/MedImageRetrieval.
| [
{
"version": "v1",
"created": "Sat, 14 Sep 2024 13:07:30 GMT"
},
{
"version": "v2",
"created": "Wed, 26 Mar 2025 19:11:03 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Mahbod",
"Amirreza",
""
],
[
"Saeidi",
"Nematollah",
""
],
[
"Hatamikia",
"Sepideh",
""
],
[
"Woitek",
"Ramona",
""
]
] | TITLE: Evaluating Pre-trained Convolutional Neural Networks and Foundation
Models as Feature Extractors for Content-based Medical Image Retrieval
ABSTRACT: Medical image retrieval refers to the task of finding similar images for
given query images in a database, with applications such as diagnosis support.
While traditional medical image retrieval relied on clinical metadata,
content-based medical image retrieval (CBMIR) depends on image features, which
can be extracted automatically or semi-automatically. Many approaches have been
proposed for CBMIR, and among them, using pre-trained convolutional neural
networks (CNNs) is a widely utilized approach. However, considering the recent
advances in the development of foundation models for various computer vision
tasks, their application for CBMIR can also be investigated.
In this study, we used several pre-trained feature extractors from well-known
pre-trained CNNs and pre-trained foundation models and investigated the CBMIR
performance on eight types of two-dimensional (2D) and three-dimensional (3D)
medical images. Furthermore, we investigated the effect of image size on the
CBMIR performance.
Our results show that, overall, for the 2D datasets, foundation models
deliver superior performance by a large margin compared to CNNs, with the
general-purpose self-supervised model for computational pathology (UNI)
providing the best overall performance across all datasets and image sizes. For
3D datasets, CNNs and foundation models deliver more competitive performance,
with contrastive learning from captions for histopathology model (CONCH)
achieving the best overall performance. Moreover, our findings confirm that
while using larger image sizes (especially for 2D datasets) yields slightly
better performance, competitive CBMIR performance can still be achieved even
with smaller image sizes. Our codes to reproduce the results are available at:
https://github.com/masih4/MedImageRetrieval.
|
2409.11593 | Xing Chen | Xing Chen, Dongshu Liu, Jeremie Laydevant, Julie Grollier | Self-Contrastive Forward-Forward Algorithm | null | null | null | null | cs.LG cs.AI cs.CV cs.ET cs.NE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Agents that operate autonomously benefit from lifelong learning capabilities.
However, compatible training algorithms must comply with the decentralized
nature of these systems, which imposes constraints on both the parameter counts
and the computational resources. The Forward-Forward (FF) algorithm is one of
these. FF relies only on feedforward operations, the same used for inference,
for optimizing layer-wise objectives. This purely forward approach eliminates
the need for transpose operations required in traditional backpropagation.
Despite its potential, FF has failed to reach state-of-the-art performance on
most standard benchmark tasks, in part due to unreliable negative data
generation methods for unsupervised learning.
In this work, we propose the Self-Contrastive Forward-Forward (SCFF)
algorithm, a competitive training method aimed at closing this performance gap.
Inspired by standard self-supervised contrastive learning for vision tasks,
SCFF generates positive and negative inputs applicable across various datasets.
The method demonstrates superior performance compared to existing unsupervised
local learning algorithms on several benchmark datasets, including MNIST,
CIFAR-10, STL-10, and Tiny ImageNet. We extend FF's application to training
recurrent neural networks, expanding its utility to sequential data tasks.
These findings pave the way for high-accuracy, real-time learning on
resource-constrained edge devices.
| [
{
"version": "v1",
"created": "Tue, 17 Sep 2024 22:58:20 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 15:57:57 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Chen",
"Xing",
""
],
[
"Liu",
"Dongshu",
""
],
[
"Laydevant",
"Jeremie",
""
],
[
"Grollier",
"Julie",
""
]
] | TITLE: Self-Contrastive Forward-Forward Algorithm
ABSTRACT: Agents that operate autonomously benefit from lifelong learning capabilities.
However, compatible training algorithms must comply with the decentralized
nature of these systems, which imposes constraints on both the parameter counts
and the computational resources. The Forward-Forward (FF) algorithm is one of
these. FF relies only on feedforward operations, the same used for inference,
for optimizing layer-wise objectives. This purely forward approach eliminates
the need for transpose operations required in traditional backpropagation.
Despite its potential, FF has failed to reach state-of-the-art performance on
most standard benchmark tasks, in part due to unreliable negative data
generation methods for unsupervised learning.
In this work, we propose the Self-Contrastive Forward-Forward (SCFF)
algorithm, a competitive training method aimed at closing this performance gap.
Inspired by standard self-supervised contrastive learning for vision tasks,
SCFF generates positive and negative inputs applicable across various datasets.
The method demonstrates superior performance compared to existing unsupervised
local learning algorithms on several benchmark datasets, including MNIST,
CIFAR-10, STL-10, and Tiny ImageNet. We extend FF's application to training
recurrent neural networks, expanding its utility to sequential data tasks.
These findings pave the way for high-accuracy, real-time learning on
resource-constrained edge devices.
|
2409.12249 | Yipeng Xu | Yuzhe Wu, Yipeng Xu, Tianyu Xu, Jialu Zhang, Jianfeng Ren, Xudong
Jiang | GCA-SUNet: A Gated Context-Aware Swin-UNet for Exemplar-Free Counting | Accepted by ICME 2025 | null | null | null | cs.CV cs.AI | http://creativecommons.org/licenses/by/4.0/ | Exemplar-Free Counting aims to count objects of interest without intensive
annotations of objects or exemplars. To achieve this, we propose a Gated
Context-Aware Swin-UNet (GCA-SUNet) to directly map an input image to the
density map of countable objects. Specifically, a set of Swin transformers form
an encoder to derive a robust feature representation, and a Gated Context-Aware
Modulation block is designed to suppress irrelevant objects or background
through a gate mechanism and exploit the attentive support of objects of
interest through a self-similarity matrix. The gate strategy is also
incorporated into the bottleneck network and the decoder of the Swin-UNet to
highlight the features most relevant to objects of interest. By explicitly
exploiting the attentive support among countable objects and eliminating
irrelevant features through the gate mechanisms, the proposed GCA-SUNet focuses
on and counts objects of interest without relying on predefined categories or
exemplars. Experimental results on the real-world datasets such as FSC-147 and
CARPK demonstrate that GCA-SUNet significantly and consistently outperforms
state-of-the-art methods. The code is available at
https://github.com/Amordia/GCA-SUNet.
| [
{
"version": "v1",
"created": "Wed, 18 Sep 2024 18:14:00 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 00:09:03 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Wu",
"Yuzhe",
""
],
[
"Xu",
"Yipeng",
""
],
[
"Xu",
"Tianyu",
""
],
[
"Zhang",
"Jialu",
""
],
[
"Ren",
"Jianfeng",
""
],
[
"Jiang",
"Xudong",
""
]
] | TITLE: GCA-SUNet: A Gated Context-Aware Swin-UNet for Exemplar-Free Counting
ABSTRACT: Exemplar-Free Counting aims to count objects of interest without intensive
annotations of objects or exemplars. To achieve this, we propose a Gated
Context-Aware Swin-UNet (GCA-SUNet) to directly map an input image to the
density map of countable objects. Specifically, a set of Swin transformers form
an encoder to derive a robust feature representation, and a Gated Context-Aware
Modulation block is designed to suppress irrelevant objects or background
through a gate mechanism and exploit the attentive support of objects of
interest through a self-similarity matrix. The gate strategy is also
incorporated into the bottleneck network and the decoder of the Swin-UNet to
highlight the features most relevant to objects of interest. By explicitly
exploiting the attentive support among countable objects and eliminating
irrelevant features through the gate mechanisms, the proposed GCA-SUNet focuses
on and counts objects of interest without relying on predefined categories or
exemplars. Experimental results on the real-world datasets such as FSC-147 and
CARPK demonstrate that GCA-SUNet significantly and consistently outperforms
state-of-the-art methods. The code is available at
https://github.com/Amordia/GCA-SUNet.
|
2409.12259 | Rolandos Alexandros Potamias | Rolandos Alexandros Potamias and Jinglei Zhang and Jiankang Deng and
Stefanos Zafeiriou | WiLoR: End-to-end 3D Hand Localization and Reconstruction in-the-wild | CVPR 2025, Project Page https://rolpotamias.github.io/WiLoR | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-nd/4.0/ | In recent years, 3D hand pose estimation methods have garnered significant
attention due to their extensive applications in human-computer interaction,
virtual reality, and robotics. In contrast, there has been a notable gap in
hand detection pipelines, posing significant challenges in constructing
effective real-world multi-hand reconstruction systems. In this work, we
present a data-driven pipeline for efficient multi-hand reconstruction in the
wild. The proposed pipeline is composed of two components: a real-time fully
convolutional hand localization and a high-fidelity transformer-based 3D hand
reconstruction model. To tackle the limitations of previous methods and build a
robust and stable detection network, we introduce a large-scale dataset with
over than 2M in-the-wild hand images with diverse lighting, illumination, and
occlusion conditions. Our approach outperforms previous methods in both
efficiency and accuracy on popular 2D and 3D benchmarks. Finally, we showcase
the effectiveness of our pipeline to achieve smooth 3D hand tracking from
monocular videos, without utilizing any temporal components. Code, models, and
dataset are available https://rolpotamias.github.io/WiLoR.
| [
{
"version": "v1",
"created": "Wed, 18 Sep 2024 18:46:51 GMT"
},
{
"version": "v2",
"created": "Wed, 26 Mar 2025 18:05:52 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Potamias",
"Rolandos Alexandros",
""
],
[
"Zhang",
"Jinglei",
""
],
[
"Deng",
"Jiankang",
""
],
[
"Zafeiriou",
"Stefanos",
""
]
] | TITLE: WiLoR: End-to-end 3D Hand Localization and Reconstruction in-the-wild
ABSTRACT: In recent years, 3D hand pose estimation methods have garnered significant
attention due to their extensive applications in human-computer interaction,
virtual reality, and robotics. In contrast, there has been a notable gap in
hand detection pipelines, posing significant challenges in constructing
effective real-world multi-hand reconstruction systems. In this work, we
present a data-driven pipeline for efficient multi-hand reconstruction in the
wild. The proposed pipeline is composed of two components: a real-time fully
convolutional hand localization and a high-fidelity transformer-based 3D hand
reconstruction model. To tackle the limitations of previous methods and build a
robust and stable detection network, we introduce a large-scale dataset with
over than 2M in-the-wild hand images with diverse lighting, illumination, and
occlusion conditions. Our approach outperforms previous methods in both
efficiency and accuracy on popular 2D and 3D benchmarks. Finally, we showcase
the effectiveness of our pipeline to achieve smooth 3D hand tracking from
monocular videos, without utilizing any temporal components. Code, models, and
dataset are available https://rolpotamias.github.io/WiLoR.
|
2409.15272 | Yizhi Li | Yizhi Li, Ge Zhang, Yinghao Ma, Ruibin Yuan, Kang Zhu, Hangyu Guo,
Yiming Liang, Jiaheng Liu, Zekun Wang, Jian Yang, Siwei Wu, Xingwei Qu,
Jinjie Shi, Xinyue Zhang, Zhenzhu Yang, Xiangzhou Wang, Zhaoxiang Zhang,
Zachary Liu, Emmanouil Benetos, Wenhao Huang, Chenghua Lin | OmniBench: Towards The Future of Universal Omni-Language Models | null | null | null | null | cs.CL cs.AI cs.CV | http://creativecommons.org/licenses/by-sa/4.0/ | Recent advancements in multimodal large language models (MLLMs) have focused
on integrating multiple modalities, yet their ability to simultaneously process
and reason across different inputs remains underexplored. We introduce
OmniBench, a novel benchmark designed to evaluate models' ability to recognize,
interpret, and reason across visual, acoustic, and textual inputs
simultaneously. We define language models capable of such tri-modal processing
as omni-language models (OLMs). OmniBench features high-quality human
annotations that require integrated understanding across all modalities. Our
evaluation reveals that: i) open-source OLMs show significant limitations in
instruction-following and reasoning in tri-modal contexts; and ii) most
baseline models perform poorly (around 50% accuracy) even with textual
alternatives to image/audio inputs. To address these limitations, we develop
OmniInstruct, an 96K-sample instruction tuning dataset for training OLMs. We
advocate for developing more robust tri-modal integration techniques and
training strategies to enhance OLM performance. Codes and data could be found
at our repo (https://github.com/multimodal-art-projection/OmniBench).
| [
{
"version": "v1",
"created": "Mon, 23 Sep 2024 17:59:05 GMT"
},
{
"version": "v2",
"created": "Tue, 24 Sep 2024 16:51:45 GMT"
},
{
"version": "v3",
"created": "Thu, 3 Oct 2024 22:32:50 GMT"
},
{
"version": "v4",
"created": "Thu, 27 Mar 2025 16:21:06 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Li",
"Yizhi",
""
],
[
"Zhang",
"Ge",
""
],
[
"Ma",
"Yinghao",
""
],
[
"Yuan",
"Ruibin",
""
],
[
"Zhu",
"Kang",
""
],
[
"Guo",
"Hangyu",
""
],
[
"Liang",
"Yiming",
""
],
[
"Liu",
"Jiaheng",
""
],
[
"Wang",
"Zekun",
""
],
[
"Yang",
"Jian",
""
],
[
"Wu",
"Siwei",
""
],
[
"Qu",
"Xingwei",
""
],
[
"Shi",
"Jinjie",
""
],
[
"Zhang",
"Xinyue",
""
],
[
"Yang",
"Zhenzhu",
""
],
[
"Wang",
"Xiangzhou",
""
],
[
"Zhang",
"Zhaoxiang",
""
],
[
"Liu",
"Zachary",
""
],
[
"Benetos",
"Emmanouil",
""
],
[
"Huang",
"Wenhao",
""
],
[
"Lin",
"Chenghua",
""
]
] | TITLE: OmniBench: Towards The Future of Universal Omni-Language Models
ABSTRACT: Recent advancements in multimodal large language models (MLLMs) have focused
on integrating multiple modalities, yet their ability to simultaneously process
and reason across different inputs remains underexplored. We introduce
OmniBench, a novel benchmark designed to evaluate models' ability to recognize,
interpret, and reason across visual, acoustic, and textual inputs
simultaneously. We define language models capable of such tri-modal processing
as omni-language models (OLMs). OmniBench features high-quality human
annotations that require integrated understanding across all modalities. Our
evaluation reveals that: i) open-source OLMs show significant limitations in
instruction-following and reasoning in tri-modal contexts; and ii) most
baseline models perform poorly (around 50% accuracy) even with textual
alternatives to image/audio inputs. To address these limitations, we develop
OmniInstruct, an 96K-sample instruction tuning dataset for training OLMs. We
advocate for developing more robust tri-modal integration techniques and
training strategies to enhance OLM performance. Codes and data could be found
at our repo (https://github.com/multimodal-art-projection/OmniBench).
|
2409.18119 | Yuexi Du | Yuexi Du, John Onofrey, Nicha C. Dvornek | Multi-View and Multi-Scale Alignment for Contrastive Language-Image
Pre-training in Mammography | This paper is accepted by IPMI 2025 for Oral Presentation | null | null | null | cs.CV cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Contrastive Language-Image Pre-training (CLIP) demonstrates strong potential
in medical image analysis but requires substantial data and computational
resources. Due to these restrictions, existing CLIP applications in medical
imaging focus mainly on modalities like chest X-rays that have abundant
image-report data available, leaving many other important modalities
underexplored. Here, we propose one of the first adaptations of the full CLIP
model to mammography, which presents significant challenges due to labeled data
scarcity, high-resolution images with small regions of interest, and class-wise
imbalance. We first develop a specialized supervision framework for mammography
that leverages its multi-view nature. Furthermore, we design a symmetric local
alignment module to better focus on detailed features in high-resolution
images. Lastly, we incorporate a parameter-efficient fine-tuning approach for
large language models pre-trained with medical knowledge to address data
limitations. Our multi-view and multi-scale alignment (MaMA) method outperforms
state-of-the-art baselines for three different tasks on two large real-world
mammography datasets, EMBED and RSNA-Mammo, with only 52% model size compared
with the largest baseline. The code is available at
https://github.com/XYPB/MaMA
| [
{
"version": "v1",
"created": "Thu, 26 Sep 2024 17:56:59 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 17:39:55 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Du",
"Yuexi",
""
],
[
"Onofrey",
"John",
""
],
[
"Dvornek",
"Nicha C.",
""
]
] | TITLE: Multi-View and Multi-Scale Alignment for Contrastive Language-Image
Pre-training in Mammography
ABSTRACT: Contrastive Language-Image Pre-training (CLIP) demonstrates strong potential
in medical image analysis but requires substantial data and computational
resources. Due to these restrictions, existing CLIP applications in medical
imaging focus mainly on modalities like chest X-rays that have abundant
image-report data available, leaving many other important modalities
underexplored. Here, we propose one of the first adaptations of the full CLIP
model to mammography, which presents significant challenges due to labeled data
scarcity, high-resolution images with small regions of interest, and class-wise
imbalance. We first develop a specialized supervision framework for mammography
that leverages its multi-view nature. Furthermore, we design a symmetric local
alignment module to better focus on detailed features in high-resolution
images. Lastly, we incorporate a parameter-efficient fine-tuning approach for
large language models pre-trained with medical knowledge to address data
limitations. Our multi-view and multi-scale alignment (MaMA) method outperforms
state-of-the-art baselines for three different tasks on two large real-world
mammography datasets, EMBED and RSNA-Mammo, with only 52% model size compared
with the largest baseline. The code is available at
https://github.com/XYPB/MaMA
|
2409.19804 | Xuyang Wu | Xuyang Wu, Shuowei Li, Hsin-Tai Wu, Zhiqiang Tao and Yi Fang | Does RAG Introduce Unfairness in LLMs? Evaluating Fairness in
Retrieval-Augmented Generation Systems | Published at COLING 2025 | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Retrieval-Augmented Generation (RAG) has recently gained significant
attention for its enhanced ability to integrate external knowledge sources into
open-domain question answering (QA) tasks. However, it remains unclear how
these models address fairness concerns, particularly with respect to sensitive
attributes such as gender, geographic location, and other demographic factors.
First, as language models evolve to prioritize utility, like improving exact
match accuracy, fairness considerations may have been largely overlooked.
Second, the complex, multi-component architecture of RAG methods poses
challenges in identifying and mitigating biases, as each component is optimized
for distinct objectives. In this paper, we aim to empirically evaluate fairness
in several RAG methods. We propose a fairness evaluation framework tailored to
RAG, using scenario-based questions and analyzing disparities across
demographic attributes. Our experimental results indicate that, despite recent
advances in utility-driven optimization, fairness issues persist in both the
retrieval and generation stages. These findings underscore the need for
targeted interventions to address fairness concerns throughout the RAG
pipeline. The dataset and code used in this study are publicly available at
this GitHub Repository https://github.com/elviswxy/RAG_fairness .
| [
{
"version": "v1",
"created": "Sun, 29 Sep 2024 22:04:26 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 04:36:46 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Wu",
"Xuyang",
""
],
[
"Li",
"Shuowei",
""
],
[
"Wu",
"Hsin-Tai",
""
],
[
"Tao",
"Zhiqiang",
""
],
[
"Fang",
"Yi",
""
]
] | TITLE: Does RAG Introduce Unfairness in LLMs? Evaluating Fairness in
Retrieval-Augmented Generation Systems
ABSTRACT: Retrieval-Augmented Generation (RAG) has recently gained significant
attention for its enhanced ability to integrate external knowledge sources into
open-domain question answering (QA) tasks. However, it remains unclear how
these models address fairness concerns, particularly with respect to sensitive
attributes such as gender, geographic location, and other demographic factors.
First, as language models evolve to prioritize utility, like improving exact
match accuracy, fairness considerations may have been largely overlooked.
Second, the complex, multi-component architecture of RAG methods poses
challenges in identifying and mitigating biases, as each component is optimized
for distinct objectives. In this paper, we aim to empirically evaluate fairness
in several RAG methods. We propose a fairness evaluation framework tailored to
RAG, using scenario-based questions and analyzing disparities across
demographic attributes. Our experimental results indicate that, despite recent
advances in utility-driven optimization, fairness issues persist in both the
retrieval and generation stages. These findings underscore the need for
targeted interventions to address fairness concerns throughout the RAG
pipeline. The dataset and code used in this study are publicly available at
this GitHub Repository https://github.com/elviswxy/RAG_fairness .
|
2410.00068 | Xinyuan Zheng | Xinyuan Zheng, Orren Ravid, Robert A.J. Barry, Yoojean Kim, Qian Wang,
Young-geun Kim, Xi Zhu, Xiaofu He | Denoising VAE as an Explainable Feature Reduction and Diagnostic
Pipeline for Autism Based on Resting state fMRI | null | null | null | null | eess.IV cs.LG stat.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Autism spectrum disorders (ASDs) are developmental conditions characterized
by restricted interests and difficulties in communication. The complexity of
ASD has resulted in a deficiency of objective diagnostic biomarkers. Deep
learning methods have gained recognition for addressing these challenges in
neuroimaging analysis, but finding and interpreting such diagnostic biomarkers
are still challenging computationally. Here, we propose a feature reduction
pipeline using resting-state fMRI data. We used Craddock atlas and Power atlas
to extract functional connectivity data from rs-fMRI, resulting in over 30
thousand features. By using a denoising variational autoencoder, our proposed
pipeline further compresses the connectivity features into 5 latent Gaussian
distributions, providing is a low-dimensional representation of the data to
promote computational efficiency and interpretability. To test the method, we
employed the extracted latent representations to classify ASD using traditional
classifiers such as SVM on a large multi-site dataset. The 95% confidence
interval for the prediction accuracy of SVM is [0.63, 0.76] after site
harmonization using the extracted latent distributions. Without using DVAE for
dimensionality reduction, the prediction accuracy is 0.70, which falls within
the interval. The DVAE successfully encoded the diagnostic information from
rs-fMRI data without sacrificing prediction performance. The runtime for
training the DVAE and obtaining classification results from its extracted
latent features was 7 times shorter compared to training classifiers directly
on the raw data. Our findings suggest that the Power atlas provides more
effective brain connectivity insights for diagnosing ASD than Craddock atlas.
Additionally, we visualized the latent representations to gain insights into
the brain networks contributing to the differences between ASD and neurotypical
brains.
| [
{
"version": "v1",
"created": "Mon, 30 Sep 2024 09:38:47 GMT"
},
{
"version": "v2",
"created": "Sun, 5 Jan 2025 21:50:03 GMT"
},
{
"version": "v3",
"created": "Thu, 27 Mar 2025 16:25:38 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Zheng",
"Xinyuan",
""
],
[
"Ravid",
"Orren",
""
],
[
"Barry",
"Robert A. J.",
""
],
[
"Kim",
"Yoojean",
""
],
[
"Wang",
"Qian",
""
],
[
"Kim",
"Young-geun",
""
],
[
"Zhu",
"Xi",
""
],
[
"He",
"Xiaofu",
""
]
] | TITLE: Denoising VAE as an Explainable Feature Reduction and Diagnostic
Pipeline for Autism Based on Resting state fMRI
ABSTRACT: Autism spectrum disorders (ASDs) are developmental conditions characterized
by restricted interests and difficulties in communication. The complexity of
ASD has resulted in a deficiency of objective diagnostic biomarkers. Deep
learning methods have gained recognition for addressing these challenges in
neuroimaging analysis, but finding and interpreting such diagnostic biomarkers
are still challenging computationally. Here, we propose a feature reduction
pipeline using resting-state fMRI data. We used Craddock atlas and Power atlas
to extract functional connectivity data from rs-fMRI, resulting in over 30
thousand features. By using a denoising variational autoencoder, our proposed
pipeline further compresses the connectivity features into 5 latent Gaussian
distributions, providing is a low-dimensional representation of the data to
promote computational efficiency and interpretability. To test the method, we
employed the extracted latent representations to classify ASD using traditional
classifiers such as SVM on a large multi-site dataset. The 95% confidence
interval for the prediction accuracy of SVM is [0.63, 0.76] after site
harmonization using the extracted latent distributions. Without using DVAE for
dimensionality reduction, the prediction accuracy is 0.70, which falls within
the interval. The DVAE successfully encoded the diagnostic information from
rs-fMRI data without sacrificing prediction performance. The runtime for
training the DVAE and obtaining classification results from its extracted
latent features was 7 times shorter compared to training classifiers directly
on the raw data. Our findings suggest that the Power atlas provides more
effective brain connectivity insights for diagnosing ASD than Craddock atlas.
Additionally, we visualized the latent representations to gain insights into
the brain networks contributing to the differences between ASD and neurotypical
brains.
|
2410.12399 | Xuyuan Li | Xuyuan Li, Zengqiang Shang, Hua Hua, Peiyang Shi, Chen Yang, Li Wang,
Pengyuan Zhang | SF-Speech: Straightened Flow for Zero-Shot Voice Clone | Accepted by IEEE Transactions on Audio, Speech and Language
Processing | null | null | null | cs.SD eess.AS | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Recently, neural ordinary differential equations (ODE) models trained with
flow matching have achieved impressive performance on the zero-shot voice clone
task. Nevertheless, postulating standard Gaussian noise as the initial
distribution of ODE gives rise to numerous intersections within the fitted
targets of flow matching, which presents challenges to model training and
enhances the curvature of the learned generated trajectories. These curved
trajectories restrict the capacity of ODE models for generating desirable
samples with a few steps. This paper proposes SF-Speech, a novel voice clone
model based on ODE and in-context learning. Unlike the previous works,
SF-Speech adopts a lightweight multi-stage module to generate a more
deterministic initial distribution for ODE. Without introducing any additional
loss function, we effectively straighten the curved reverse trajectories of the
ODE model by jointly training it with the proposed module. Experiment results
on datasets of various scales show that SF-Speech outperforms the
state-of-the-art zero-shot TTS methods and requires only a quarter of the
solver steps, resulting in a generation speed approximately 3.7 times that of
Voicebox and E2 TTS. Audio samples are available at the demo
page\footnote{[Online] Available: https://lixuyuan102.github.io/Demo/}.
| [
{
"version": "v1",
"created": "Wed, 16 Oct 2024 09:27:25 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 13:14:57 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Li",
"Xuyuan",
""
],
[
"Shang",
"Zengqiang",
""
],
[
"Hua",
"Hua",
""
],
[
"Shi",
"Peiyang",
""
],
[
"Yang",
"Chen",
""
],
[
"Wang",
"Li",
""
],
[
"Zhang",
"Pengyuan",
""
]
] | TITLE: SF-Speech: Straightened Flow for Zero-Shot Voice Clone
ABSTRACT: Recently, neural ordinary differential equations (ODE) models trained with
flow matching have achieved impressive performance on the zero-shot voice clone
task. Nevertheless, postulating standard Gaussian noise as the initial
distribution of ODE gives rise to numerous intersections within the fitted
targets of flow matching, which presents challenges to model training and
enhances the curvature of the learned generated trajectories. These curved
trajectories restrict the capacity of ODE models for generating desirable
samples with a few steps. This paper proposes SF-Speech, a novel voice clone
model based on ODE and in-context learning. Unlike the previous works,
SF-Speech adopts a lightweight multi-stage module to generate a more
deterministic initial distribution for ODE. Without introducing any additional
loss function, we effectively straighten the curved reverse trajectories of the
ODE model by jointly training it with the proposed module. Experiment results
on datasets of various scales show that SF-Speech outperforms the
state-of-the-art zero-shot TTS methods and requires only a quarter of the
solver steps, resulting in a generation speed approximately 3.7 times that of
Voicebox and E2 TTS. Audio samples are available at the demo
page\footnote{[Online] Available: https://lixuyuan102.github.io/Demo/}.
|
2410.14379 | Ziming Huang | Ziming Huang, Xurui Li, Haotian Liu, Feng Xue, Yuzhe Wang, Yu Zhou | AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial
Scenarios | Accepted at CVPR2025 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recently, multi-class anomaly classification has garnered increasing
attention. Previous methods directly cluster anomalies but often struggle due
to the lack of anomaly-prior knowledge. Acquiring this knowledge faces two
issues: the non-prominent and weak-semantics anomalies. In this paper, we
propose AnomalyNCD, a multi-class anomaly classification network compatible
with different anomaly detection methods. To address the non-prominence of
anomalies, we design main element binarization (MEBin) to obtain
anomaly-centered images, ensuring anomalies are learned while avoiding the
impact of incorrect detections. Next, to learn anomalies with weak semantics,
we design mask-guided representation learning, which focuses on isolated
anomalies guided by masks and reduces confusion from erroneous inputs through
corrected pseudo labels. Finally, to enable flexible classification at both
region and image levels, we develop a region merging strategy that determines
the overall image category based on the classified anomaly regions. Our method
outperforms the state-of-the-art works on the MVTec AD and MTD datasets.
Compared with the current methods, AnomalyNCD combined with zero-shot anomaly
detection method achieves a 10.8% $F_1$ gain, 8.8% NMI gain, and 9.5% ARI gain
on MVTec AD, and 12.8% $F_1$ gain, 5.7% NMI gain, and 10.8% ARI gain on MTD.
Code is available at https://github.com/HUST-SLOW/AnomalyNCD.
| [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 11:07:12 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 13:09:07 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Huang",
"Ziming",
""
],
[
"Li",
"Xurui",
""
],
[
"Liu",
"Haotian",
""
],
[
"Xue",
"Feng",
""
],
[
"Wang",
"Yuzhe",
""
],
[
"Zhou",
"Yu",
""
]
] | TITLE: AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial
Scenarios
ABSTRACT: Recently, multi-class anomaly classification has garnered increasing
attention. Previous methods directly cluster anomalies but often struggle due
to the lack of anomaly-prior knowledge. Acquiring this knowledge faces two
issues: the non-prominent and weak-semantics anomalies. In this paper, we
propose AnomalyNCD, a multi-class anomaly classification network compatible
with different anomaly detection methods. To address the non-prominence of
anomalies, we design main element binarization (MEBin) to obtain
anomaly-centered images, ensuring anomalies are learned while avoiding the
impact of incorrect detections. Next, to learn anomalies with weak semantics,
we design mask-guided representation learning, which focuses on isolated
anomalies guided by masks and reduces confusion from erroneous inputs through
corrected pseudo labels. Finally, to enable flexible classification at both
region and image levels, we develop a region merging strategy that determines
the overall image category based on the classified anomaly regions. Our method
outperforms the state-of-the-art works on the MVTec AD and MTD datasets.
Compared with the current methods, AnomalyNCD combined with zero-shot anomaly
detection method achieves a 10.8% $F_1$ gain, 8.8% NMI gain, and 9.5% ARI gain
on MVTec AD, and 12.8% $F_1$ gain, 5.7% NMI gain, and 10.8% ARI gain on MTD.
Code is available at https://github.com/HUST-SLOW/AnomalyNCD.
|
2410.14770 | Jiaxin Lu | Jiaxin Lu, Yongqing Liang, Huijun Han, Jiacheng Hua, Junfeng Jiang,
Xin Li, Qixing Huang | A Survey on Computational Solutions for Reconstructing Complete Objects
by Reassembling Their Fractured Parts | 36 pages, 22 figures | null | null | null | cs.CV cs.GR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Reconstructing a complete object from its parts is a fundamental problem in
many scientific domains. The purpose of this article is to provide a systematic
survey on this topic. The reassembly problem requires understanding the
attributes of individual pieces and establishing matches between different
pieces. Many approaches also model priors of the underlying complete object.
Existing approaches are tightly connected problems of shape segmentation, shape
matching, and learning shape priors. We provide existing algorithms in this
context and emphasize their similarities and differences to general-purpose
approaches. We also survey the trends from early non-deep learning approaches
to more recent deep learning approaches. In addition to algorithms, this survey
will also describe existing datasets, open-source software packages, and
applications. To the best of our knowledge, this is the first comprehensive
survey on this topic in computer graphics.
| [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 17:53:07 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 17:45:43 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Lu",
"Jiaxin",
""
],
[
"Liang",
"Yongqing",
""
],
[
"Han",
"Huijun",
""
],
[
"Hua",
"Jiacheng",
""
],
[
"Jiang",
"Junfeng",
""
],
[
"Li",
"Xin",
""
],
[
"Huang",
"Qixing",
""
]
] | TITLE: A Survey on Computational Solutions for Reconstructing Complete Objects
by Reassembling Their Fractured Parts
ABSTRACT: Reconstructing a complete object from its parts is a fundamental problem in
many scientific domains. The purpose of this article is to provide a systematic
survey on this topic. The reassembly problem requires understanding the
attributes of individual pieces and establishing matches between different
pieces. Many approaches also model priors of the underlying complete object.
Existing approaches are tightly connected problems of shape segmentation, shape
matching, and learning shape priors. We provide existing algorithms in this
context and emphasize their similarities and differences to general-purpose
approaches. We also survey the trends from early non-deep learning approaches
to more recent deep learning approaches. In addition to algorithms, this survey
will also describe existing datasets, open-source software packages, and
applications. To the best of our knowledge, this is the first comprehensive
survey on this topic in computer graphics.
|
2410.21897 | Monan Zhou Dr | Yifu Sun, Xulong Zhang, Monan Zhou, Wei Li | Semi-Supervised Self-Learning Enhanced Music Emotion Recognition | 12 pages, 2 figures | Proceedings of the 11th Conference on Sound and Music Technology.
CSMT 2024. Lecture Notes in Electrical Engineering. Springer, Singapore | null | null | cs.SD cs.AI eess.AS | http://creativecommons.org/licenses/by/4.0/ | Music emotion recognition (MER) aims to identify the emotions conveyed in a
given musical piece. However, currently, in the field of MER, the available
public datasets have limited sample sizes. Recently, segment-based methods for
emotion-related tasks have been proposed, which train backbone networks on
shorter segments instead of entire audio clips, thereby naturally augmenting
training samples without requiring additional resources. Then, the predicted
segment-level results are aggregated to obtain the entire song prediction. The
most commonly used method is that the segment inherits the label of the clip
containing it, but music emotion is not constant during the whole clip. Doing
so will introduce label noise and make the training easy to overfit. To handle
the noisy label issue, we propose a semi-supervised self-learning (SSSL)
method, which can differentiate between samples with correct and incorrect
labels in a self-learning manner, thus effectively utilizing the augmented
segment-level data. Experiments on three public emotional datasets demonstrate
that the proposed method can achieve better or comparable performance.
| [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 09:42:07 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 02:39:50 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Sun",
"Yifu",
""
],
[
"Zhang",
"Xulong",
""
],
[
"Zhou",
"Monan",
""
],
[
"Li",
"Wei",
""
]
] | TITLE: Semi-Supervised Self-Learning Enhanced Music Emotion Recognition
ABSTRACT: Music emotion recognition (MER) aims to identify the emotions conveyed in a
given musical piece. However, currently, in the field of MER, the available
public datasets have limited sample sizes. Recently, segment-based methods for
emotion-related tasks have been proposed, which train backbone networks on
shorter segments instead of entire audio clips, thereby naturally augmenting
training samples without requiring additional resources. Then, the predicted
segment-level results are aggregated to obtain the entire song prediction. The
most commonly used method is that the segment inherits the label of the clip
containing it, but music emotion is not constant during the whole clip. Doing
so will introduce label noise and make the training easy to overfit. To handle
the noisy label issue, we propose a semi-supervised self-learning (SSSL)
method, which can differentiate between samples with correct and incorrect
labels in a self-learning manner, thus effectively utilizing the augmented
segment-level data. Experiments on three public emotional datasets demonstrate
that the proposed method can achieve better or comparable performance.
|
2411.01739 | Yanyi Zhang | Yanyi Zhang, Binglin Qiu, Qi Jia, Yu Liu, Ran He | Not Just Object, But State: Compositional Incremental Learning without
Forgetting | NeurIPS 2024 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Most incremental learners excessively prioritize coarse classes of objects
while neglecting various kinds of states (e.g. color and material) attached to
the objects. As a result, they are limited in the ability to reason
fine-grained compositionality of state-object pairs. To remedy this limitation,
we propose a novel task called Compositional Incremental Learning
(composition-IL), enabling the model to recognize state-object compositions as
a whole in an incremental learning fashion. Since the lack of suitable
benchmarks, we re-organize two existing datasets and make them tailored for
composition-IL. Then, we propose a prompt-based Composition Incremental Learner
(CompILer), to overcome the ambiguous composition boundary problem which
challenges composition-IL largely. Specifically, we exploit multi-pool prompt
learning, which is regularized by inter-pool prompt discrepancy and intra-pool
prompt diversity. Besides, we devise object-injected state prompting by using
object prompts to guide the selection of state prompts. Furthermore, we fuse
the selected prompts by a generalized-mean strategy, to eliminate irrelevant
information learned in the prompts. Extensive experiments on two datasets
exhibit state-of-the-art performance achieved by CompILer.
| [
{
"version": "v1",
"created": "Mon, 4 Nov 2024 01:42:41 GMT"
},
{
"version": "v2",
"created": "Tue, 5 Nov 2024 10:23:00 GMT"
},
{
"version": "v3",
"created": "Thu, 27 Mar 2025 08:15:46 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Zhang",
"Yanyi",
""
],
[
"Qiu",
"Binglin",
""
],
[
"Jia",
"Qi",
""
],
[
"Liu",
"Yu",
""
],
[
"He",
"Ran",
""
]
] | TITLE: Not Just Object, But State: Compositional Incremental Learning without
Forgetting
ABSTRACT: Most incremental learners excessively prioritize coarse classes of objects
while neglecting various kinds of states (e.g. color and material) attached to
the objects. As a result, they are limited in the ability to reason
fine-grained compositionality of state-object pairs. To remedy this limitation,
we propose a novel task called Compositional Incremental Learning
(composition-IL), enabling the model to recognize state-object compositions as
a whole in an incremental learning fashion. Since the lack of suitable
benchmarks, we re-organize two existing datasets and make them tailored for
composition-IL. Then, we propose a prompt-based Composition Incremental Learner
(CompILer), to overcome the ambiguous composition boundary problem which
challenges composition-IL largely. Specifically, we exploit multi-pool prompt
learning, which is regularized by inter-pool prompt discrepancy and intra-pool
prompt diversity. Besides, we devise object-injected state prompting by using
object prompts to guide the selection of state prompts. Furthermore, we fuse
the selected prompts by a generalized-mean strategy, to eliminate irrelevant
information learned in the prompts. Extensive experiments on two datasets
exhibit state-of-the-art performance achieved by CompILer.
|
2411.03055 | Luca Zhou | Luca Zhou, Daniele Solombrino, Donato Crisostomi, Maria Sofia
Bucarelli, Fabrizio Silvestri, Emanuele Rodol\`a | ATM: Improving Model Merging by Alternating Tuning and Merging | Main paper: 9 Pages, 9 figures, 1 table | null | null | null | cs.LG cs.AI cs.CV | http://creativecommons.org/licenses/by/4.0/ | Model merging has recently emerged as a cost-efficient paradigm for
multi-task learning. Among current approaches, task arithmetic stands out for
its simplicity and effectiveness. In this paper, we motivate the effectiveness
of task vectors by linking them to multi-task gradients. We show that in a
single-epoch scenario, if the optimization is performed via gradient descent,
task vectors are after one step mathematically equivalent to the gradients
obtained via gradient descent in a multi-task setting, and still approximate
these gradients in subsequent epochs. Furthermore, we show that the
effectiveness of task vectors is largely driven by the first epoch's gradient.
Given this parallel between task vectors and gradients, we propose viewing
model merging as a single step in an iterative process that alternates between
tuning and merging (ATM). We then propose two ways to utilize ATM. The first is
to replace multi-task learning with ATM in scenarios where data sharing is
prohibited, such as federated learning. The second is to improve the outcome of
any model merging algorithm by applying a few post-hoc iterations of ATM on a
small validation dataset, which is commonly available for hyperparameter
tuning. Finally, we provide both empirical and theoretical support for the
effectiveness of ATM, demonstrating that it minimizes an upper bound on the
loss obtained by jointly finetuning all tasks.
| [
{
"version": "v1",
"created": "Tue, 5 Nov 2024 12:42:42 GMT"
},
{
"version": "v2",
"created": "Wed, 6 Nov 2024 13:24:10 GMT"
},
{
"version": "v3",
"created": "Thu, 27 Mar 2025 08:57:30 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Zhou",
"Luca",
""
],
[
"Solombrino",
"Daniele",
""
],
[
"Crisostomi",
"Donato",
""
],
[
"Bucarelli",
"Maria Sofia",
""
],
[
"Silvestri",
"Fabrizio",
""
],
[
"Rodolà",
"Emanuele",
""
]
] | TITLE: ATM: Improving Model Merging by Alternating Tuning and Merging
ABSTRACT: Model merging has recently emerged as a cost-efficient paradigm for
multi-task learning. Among current approaches, task arithmetic stands out for
its simplicity and effectiveness. In this paper, we motivate the effectiveness
of task vectors by linking them to multi-task gradients. We show that in a
single-epoch scenario, if the optimization is performed via gradient descent,
task vectors are after one step mathematically equivalent to the gradients
obtained via gradient descent in a multi-task setting, and still approximate
these gradients in subsequent epochs. Furthermore, we show that the
effectiveness of task vectors is largely driven by the first epoch's gradient.
Given this parallel between task vectors and gradients, we propose viewing
model merging as a single step in an iterative process that alternates between
tuning and merging (ATM). We then propose two ways to utilize ATM. The first is
to replace multi-task learning with ATM in scenarios where data sharing is
prohibited, such as federated learning. The second is to improve the outcome of
any model merging algorithm by applying a few post-hoc iterations of ATM on a
small validation dataset, which is commonly available for hyperparameter
tuning. Finally, we provide both empirical and theoretical support for the
effectiveness of ATM, demonstrating that it minimizes an upper bound on the
loss obtained by jointly finetuning all tasks.
|
2411.04844 | Shaokai Wu | Shaokai Wu, Yuxiang Lu, Wei Ji, Suizhi Huang, Fengyu Yang, Shalayiding
Sirejiding, Qichen He, Jing Tong, Yanbiao Ji, Yue Ding, Hongtao Lu | Discretized Gaussian Representation for Tomographic Reconstruction | null | null | null | null | eess.IV cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Computed Tomography (CT) is a widely used imaging technique that provides
detailed cross-sectional views of objects. Over the past decade, Deep
Learning-based Reconstruction (DLR) methods have led efforts to enhance image
quality and reduce noise, yet they often require large amounts of data and are
computationally intensive. Inspired by recent advancements in scene
reconstruction, some approaches have adapted NeRF and 3D Gaussian Splatting
(3DGS) techniques for CT reconstruction. However, these methods are not ideal
for direct 3D volume reconstruction. In this paper, we propose a novel
Discretized Gaussian Representation (DGR) for CT reconstruction, which directly
reconstructs the 3D volume using a set of discretized Gaussian functions in an
end-to-end manner. To further enhance computational efficiency, we introduce a
Fast Volume Reconstruction technique that aggregates the contributions of these
Gaussians into a discretized volume in a highly parallelized fashion. Our
extensive experiments on both real-world and synthetic datasets demonstrate
that DGR achieves superior reconstruction quality and significantly improved
computational efficiency compared to existing DLR and instance reconstruction
methods. Our code has been provided for review purposes and will be made
publicly available upon publication.
| [
{
"version": "v1",
"created": "Thu, 7 Nov 2024 16:32:29 GMT"
},
{
"version": "v2",
"created": "Wed, 11 Dec 2024 17:40:32 GMT"
},
{
"version": "v3",
"created": "Thu, 27 Mar 2025 15:00:57 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Wu",
"Shaokai",
""
],
[
"Lu",
"Yuxiang",
""
],
[
"Ji",
"Wei",
""
],
[
"Huang",
"Suizhi",
""
],
[
"Yang",
"Fengyu",
""
],
[
"Sirejiding",
"Shalayiding",
""
],
[
"He",
"Qichen",
""
],
[
"Tong",
"Jing",
""
],
[
"Ji",
"Yanbiao",
""
],
[
"Ding",
"Yue",
""
],
[
"Lu",
"Hongtao",
""
]
] | TITLE: Discretized Gaussian Representation for Tomographic Reconstruction
ABSTRACT: Computed Tomography (CT) is a widely used imaging technique that provides
detailed cross-sectional views of objects. Over the past decade, Deep
Learning-based Reconstruction (DLR) methods have led efforts to enhance image
quality and reduce noise, yet they often require large amounts of data and are
computationally intensive. Inspired by recent advancements in scene
reconstruction, some approaches have adapted NeRF and 3D Gaussian Splatting
(3DGS) techniques for CT reconstruction. However, these methods are not ideal
for direct 3D volume reconstruction. In this paper, we propose a novel
Discretized Gaussian Representation (DGR) for CT reconstruction, which directly
reconstructs the 3D volume using a set of discretized Gaussian functions in an
end-to-end manner. To further enhance computational efficiency, we introduce a
Fast Volume Reconstruction technique that aggregates the contributions of these
Gaussians into a discretized volume in a highly parallelized fashion. Our
extensive experiments on both real-world and synthetic datasets demonstrate
that DGR achieves superior reconstruction quality and significantly improved
computational efficiency compared to existing DLR and instance reconstruction
methods. Our code has been provided for review purposes and will be made
publicly available upon publication.
|
2411.10684 | Haoxu Huang | Haoxu Huang, Cem M. Deniz, Kyunghyun Cho, Sumit Chopra, Divyam Madaan | HIST-AID: Leveraging Historical Patient Reports for Enhanced Multi-Modal
Automatic Diagnosis | In Proceedings of Machine Learning for Health | PMLR 259(2025):502-523 | null | null | eess.IV cs.CV cs.LG | http://creativecommons.org/licenses/by/4.0/ | Chest X-ray imaging is a widely accessible and non-invasive diagnostic tool
for detecting thoracic abnormalities. While numerous AI models assist
radiologists in interpreting these images, most overlook patients' historical
data. To bridge this gap, we introduce Temporal MIMIC dataset, which integrates
five years of patient history, including radiographic scans and reports from
MIMIC-CXR and MIMIC-IV, encompassing 12,221 patients and thirteen pathologies.
Building on this, we present HIST-AID, a framework that enhances automatic
diagnostic accuracy using historical reports. HIST-AID emulates the
radiologist's comprehensive approach, leveraging historical data to improve
diagnostic accuracy. Our experiments demonstrate significant improvements, with
AUROC increasing by 6.56% and AUPRC by 9.51% compared to models that rely
solely on radiographic scans. These gains were consistently observed across
diverse demographic groups, including variations in gender, age, and racial
categories. We show that while recent data boost performance, older data may
reduce accuracy due to changes in patient conditions. Our work paves the
potential of incorporating historical data for more reliable automatic
diagnosis, providing critical support for clinical decision-making.
| [
{
"version": "v1",
"created": "Sat, 16 Nov 2024 03:20:53 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Huang",
"Haoxu",
""
],
[
"Deniz",
"Cem M.",
""
],
[
"Cho",
"Kyunghyun",
""
],
[
"Chopra",
"Sumit",
""
],
[
"Madaan",
"Divyam",
""
]
] | TITLE: HIST-AID: Leveraging Historical Patient Reports for Enhanced Multi-Modal
Automatic Diagnosis
ABSTRACT: Chest X-ray imaging is a widely accessible and non-invasive diagnostic tool
for detecting thoracic abnormalities. While numerous AI models assist
radiologists in interpreting these images, most overlook patients' historical
data. To bridge this gap, we introduce Temporal MIMIC dataset, which integrates
five years of patient history, including radiographic scans and reports from
MIMIC-CXR and MIMIC-IV, encompassing 12,221 patients and thirteen pathologies.
Building on this, we present HIST-AID, a framework that enhances automatic
diagnostic accuracy using historical reports. HIST-AID emulates the
radiologist's comprehensive approach, leveraging historical data to improve
diagnostic accuracy. Our experiments demonstrate significant improvements, with
AUROC increasing by 6.56% and AUPRC by 9.51% compared to models that rely
solely on radiographic scans. These gains were consistently observed across
diverse demographic groups, including variations in gender, age, and racial
categories. We show that while recent data boost performance, older data may
reduce accuracy due to changes in patient conditions. Our work paves the
potential of incorporating historical data for more reliable automatic
diagnosis, providing critical support for clinical decision-making.
|
2411.14522 | Tianbin Li | Tianbin Li, Yanzhou Su, Wei Li, Bin Fu, Zhe Chen, Ziyan Huang, Guoan
Wang, Chenglong Ma, Ying Chen, Ming Hu, Yanjun Li, Pengcheng Chen, Xiaowei
Hu, Zhongying Deng, Yuanfeng Ji, Jin Ye, Yu Qiao, Junjun He | GMAI-VL & GMAI-VL-5.5M: A Large Vision-Language Model and A
Comprehensive Multimodal Dataset Towards General Medical AI | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Despite significant advancements in general AI, its effectiveness in the
medical domain is limited by the lack of specialized medical knowledge. To
address this, we formulate GMAI-VL-5.5M, a multimodal medical dataset created
by converting hundreds of specialized medical datasets with various annotations
into high-quality image-text pairs. This dataset offers comprehensive task
coverage, diverse modalities, and rich image-text data. Building upon this
dataset, we develop GMAI-VL, a general medical vision-language model, with a
three-stage training strategy that enhances the integration of visual and
textual information. This approach significantly improves the model's ability
to process multimodal data, supporting accurate diagnoses and clinical
decision-making. Experiments show that GMAI-VL achieves state-of-the-art
performance across various multimodal medical tasks, including visual question
answering and medical image diagnosis.
| [
{
"version": "v1",
"created": "Thu, 21 Nov 2024 18:59:36 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 15:24:29 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Li",
"Tianbin",
""
],
[
"Su",
"Yanzhou",
""
],
[
"Li",
"Wei",
""
],
[
"Fu",
"Bin",
""
],
[
"Chen",
"Zhe",
""
],
[
"Huang",
"Ziyan",
""
],
[
"Wang",
"Guoan",
""
],
[
"Ma",
"Chenglong",
""
],
[
"Chen",
"Ying",
""
],
[
"Hu",
"Ming",
""
],
[
"Li",
"Yanjun",
""
],
[
"Chen",
"Pengcheng",
""
],
[
"Hu",
"Xiaowei",
""
],
[
"Deng",
"Zhongying",
""
],
[
"Ji",
"Yuanfeng",
""
],
[
"Ye",
"Jin",
""
],
[
"Qiao",
"Yu",
""
],
[
"He",
"Junjun",
""
]
] | TITLE: GMAI-VL & GMAI-VL-5.5M: A Large Vision-Language Model and A
Comprehensive Multimodal Dataset Towards General Medical AI
ABSTRACT: Despite significant advancements in general AI, its effectiveness in the
medical domain is limited by the lack of specialized medical knowledge. To
address this, we formulate GMAI-VL-5.5M, a multimodal medical dataset created
by converting hundreds of specialized medical datasets with various annotations
into high-quality image-text pairs. This dataset offers comprehensive task
coverage, diverse modalities, and rich image-text data. Building upon this
dataset, we develop GMAI-VL, a general medical vision-language model, with a
three-stage training strategy that enhances the integration of visual and
textual information. This approach significantly improves the model's ability
to process multimodal data, supporting accurate diagnoses and clinical
decision-making. Experiments show that GMAI-VL achieves state-of-the-art
performance across various multimodal medical tasks, including visual question
answering and medical image diagnosis.
|
2411.15482 | Su Sun | Su Sun, Cheng Zhao, Zhuoyang Sun, Yingjie Victor Chen, Mei Chen | SplatFlow: Self-Supervised Dynamic Gaussian Splatting in Neural Motion
Flow Field for Autonomous Driving | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Most existing Dynamic Gaussian Splatting methods for complex dynamic urban
scenarios rely on accurate object-level supervision from expensive manual
labeling, limiting their scalability in real-world applications. In this paper,
we introduce SplatFlow, a Self-Supervised Dynamic Gaussian Splatting within
Neural Motion Flow Fields (NMFF) to learn 4D space-time representations without
requiring tracked 3D bounding boxes, enabling accurate dynamic scene
reconstruction and novel view RGB/depth/flow synthesis. SplatFlow designs a
unified framework to seamlessly integrate time-dependent 4D Gaussian
representation within NMFF, where NMFF is a set of implicit functions to model
temporal motions of both LiDAR points and Gaussians as continuous motion flow
fields. Leveraging NMFF, SplatFlow effectively decomposes static background and
dynamic objects, representing them with 3D and 4D Gaussian primitives,
respectively. NMFF also models the correspondences of each 4D Gaussian across
time, which aggregates temporal features to enhance cross-view consistency of
dynamic components. SplatFlow further improves dynamic object identification by
distilling features from 2D foundation models into 4D space-time
representation. Comprehensive evaluations conducted on the Waymo and KITTI
Datasets validate SplatFlow's state-of-the-art (SOTA) performance for both
image reconstruction and novel view synthesis in dynamic urban scenarios.
| [
{
"version": "v1",
"created": "Sat, 23 Nov 2024 07:39:30 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 00:51:33 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Sun",
"Su",
""
],
[
"Zhao",
"Cheng",
""
],
[
"Sun",
"Zhuoyang",
""
],
[
"Chen",
"Yingjie Victor",
""
],
[
"Chen",
"Mei",
""
]
] | TITLE: SplatFlow: Self-Supervised Dynamic Gaussian Splatting in Neural Motion
Flow Field for Autonomous Driving
ABSTRACT: Most existing Dynamic Gaussian Splatting methods for complex dynamic urban
scenarios rely on accurate object-level supervision from expensive manual
labeling, limiting their scalability in real-world applications. In this paper,
we introduce SplatFlow, a Self-Supervised Dynamic Gaussian Splatting within
Neural Motion Flow Fields (NMFF) to learn 4D space-time representations without
requiring tracked 3D bounding boxes, enabling accurate dynamic scene
reconstruction and novel view RGB/depth/flow synthesis. SplatFlow designs a
unified framework to seamlessly integrate time-dependent 4D Gaussian
representation within NMFF, where NMFF is a set of implicit functions to model
temporal motions of both LiDAR points and Gaussians as continuous motion flow
fields. Leveraging NMFF, SplatFlow effectively decomposes static background and
dynamic objects, representing them with 3D and 4D Gaussian primitives,
respectively. NMFF also models the correspondences of each 4D Gaussian across
time, which aggregates temporal features to enhance cross-view consistency of
dynamic components. SplatFlow further improves dynamic object identification by
distilling features from 2D foundation models into 4D space-time
representation. Comprehensive evaluations conducted on the Waymo and KITTI
Datasets validate SplatFlow's state-of-the-art (SOTA) performance for both
image reconstruction and novel view synthesis in dynamic urban scenarios.
|
2411.18620 | Zhi Zhang | Zhi Zhang, Srishti Yadav, Fengze Han, Ekaterina Shutova | Cross-modal Information Flow in Multimodal Large Language Models | null | CVPR2025 | null | null | cs.AI cs.CL cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The recent advancements in auto-regressive multimodal large language models
(MLLMs) have demonstrated promising progress for vision-language tasks. While
there exists a variety of studies investigating the processing of linguistic
information within large language models, little is currently known about the
inner working mechanism of MLLMs and how linguistic and visual information
interact within these models. In this study, we aim to fill this gap by
examining the information flow between different modalities -- language and
vision -- in MLLMs, focusing on visual question answering. Specifically, given
an image-question pair as input, we investigate where in the model and how the
visual and linguistic information are combined to generate the final
prediction. Conducting experiments with a series of models from the LLaVA
series, we find that there are two distinct stages in the process of
integration of the two modalities. In the lower layers, the model first
transfers the more general visual features of the whole image into the
representations of (linguistic) question tokens. In the middle layers, it once
again transfers visual information about specific objects relevant to the
question to the respective token positions of the question. Finally, in the
higher layers, the resulting multimodal representation is propagated to the
last position of the input sequence for the final prediction. Overall, our
findings provide a new and comprehensive perspective on the spatial and
functional aspects of image and language processing in the MLLMs, thereby
facilitating future research into multimodal information localization and
editing. Our code and collected dataset are released here:
https://github.com/FightingFighting/cross-modal-information-flow-in-MLLM.git.
| [
{
"version": "v1",
"created": "Wed, 27 Nov 2024 18:59:26 GMT"
},
{
"version": "v2",
"created": "Tue, 25 Mar 2025 18:59:50 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Zhang",
"Zhi",
""
],
[
"Yadav",
"Srishti",
""
],
[
"Han",
"Fengze",
""
],
[
"Shutova",
"Ekaterina",
""
]
] | TITLE: Cross-modal Information Flow in Multimodal Large Language Models
ABSTRACT: The recent advancements in auto-regressive multimodal large language models
(MLLMs) have demonstrated promising progress for vision-language tasks. While
there exists a variety of studies investigating the processing of linguistic
information within large language models, little is currently known about the
inner working mechanism of MLLMs and how linguistic and visual information
interact within these models. In this study, we aim to fill this gap by
examining the information flow between different modalities -- language and
vision -- in MLLMs, focusing on visual question answering. Specifically, given
an image-question pair as input, we investigate where in the model and how the
visual and linguistic information are combined to generate the final
prediction. Conducting experiments with a series of models from the LLaVA
series, we find that there are two distinct stages in the process of
integration of the two modalities. In the lower layers, the model first
transfers the more general visual features of the whole image into the
representations of (linguistic) question tokens. In the middle layers, it once
again transfers visual information about specific objects relevant to the
question to the respective token positions of the question. Finally, in the
higher layers, the resulting multimodal representation is propagated to the
last position of the input sequence for the final prediction. Overall, our
findings provide a new and comprehensive perspective on the spatial and
functional aspects of image and language processing in the MLLMs, thereby
facilitating future research into multimodal information localization and
editing. Our code and collected dataset are released here:
https://github.com/FightingFighting/cross-modal-information-flow-in-MLLM.git.
|
2411.19835 | Mario Koddenbrock | S\"onke Tenckhoff, Mario Koddenbrock, Erik Rodner | Feedback-driven object detection and iterative model improvement | Code: https://github.com/ml-lab-htw/iterative-annotate Video:
https://www.youtube.com/watch?v=CM9uhE8NN5E | https://www.gfai.de/fileadmin/Downloads/Tagungsband/gfai-tagungsband-2024.pdf | null | null | cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Automated object detection has become increasingly valuable across diverse
applications, yet efficient, high-quality annotation remains a persistent
challenge. In this paper, we present the development and evaluation of a
platform designed to interactively improve object detection models. The
platform allows uploading and annotating images as well as fine-tuning object
detection models. Users can then manually review and refine annotations,
further creating improved snapshots that are used for automatic object
detection on subsequent image uploads - a process we refer to as semi-automatic
annotation resulting in a significant gain in annotation efficiency.
Whereas iterative refinement of model results to speed up annotation has
become common practice, we are the first to quantitatively evaluate its
benefits with respect to time, effort, and interaction savings. Our
experimental results show clear evidence for a significant time reduction of up
to 53% for semi-automatic compared to manual annotation. Importantly, these
efficiency gains did not compromise annotation quality, while matching or
occasionally even exceeding the accuracy of manual annotations. These findings
demonstrate the potential of our lightweight annotation platform for creating
high-quality object detection datasets and provide best practices to guide
future development of annotation platforms.
The platform is open-source, with the frontend and backend repositories
available on GitHub. To support the understanding of our labeling process, we
have created an explanatory video demonstrating the methodology using
microscopy images of E. coli bacteria as an example.
| [
{
"version": "v1",
"created": "Fri, 29 Nov 2024 16:45:25 GMT"
},
{
"version": "v2",
"created": "Tue, 14 Jan 2025 14:53:10 GMT"
},
{
"version": "v3",
"created": "Thu, 27 Mar 2025 08:34:04 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Tenckhoff",
"Sönke",
""
],
[
"Koddenbrock",
"Mario",
""
],
[
"Rodner",
"Erik",
""
]
] | TITLE: Feedback-driven object detection and iterative model improvement
ABSTRACT: Automated object detection has become increasingly valuable across diverse
applications, yet efficient, high-quality annotation remains a persistent
challenge. In this paper, we present the development and evaluation of a
platform designed to interactively improve object detection models. The
platform allows uploading and annotating images as well as fine-tuning object
detection models. Users can then manually review and refine annotations,
further creating improved snapshots that are used for automatic object
detection on subsequent image uploads - a process we refer to as semi-automatic
annotation resulting in a significant gain in annotation efficiency.
Whereas iterative refinement of model results to speed up annotation has
become common practice, we are the first to quantitatively evaluate its
benefits with respect to time, effort, and interaction savings. Our
experimental results show clear evidence for a significant time reduction of up
to 53% for semi-automatic compared to manual annotation. Importantly, these
efficiency gains did not compromise annotation quality, while matching or
occasionally even exceeding the accuracy of manual annotations. These findings
demonstrate the potential of our lightweight annotation platform for creating
high-quality object detection datasets and provide best practices to guide
future development of annotation platforms.
The platform is open-source, with the frontend and backend repositories
available on GitHub. To support the understanding of our labeling process, we
have created an explanatory video demonstrating the methodology using
microscopy images of E. coli bacteria as an example.
|
2412.00692 | Yizhou Wang | Yizhou Wang, Tim Meinhardt, Orcun Cetintas, Cheng-Yen Yang, Sameer
Satish Pusegaonkar, Benjamin Missaoui, Sujit Biswas, Zheng Tang, Laura
Leal-Taix\'e | MCBLT: Multi-Camera Multi-Object 3D Tracking in Long Videos | null | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Object perception from multi-view cameras is crucial for intelligent systems,
particularly in indoor environments, e.g., warehouses, retail stores, and
hospitals. Most traditional multi-target multi-camera (MTMC) detection and
tracking methods rely on 2D object detection, single-view multi-object tracking
(MOT), and cross-view re-identification (ReID) techniques, without properly
handling important 3D information by multi-view image aggregation. In this
paper, we propose a 3D object detection and tracking framework, named MCBLT,
which first aggregates multi-view images with necessary camera calibration
parameters to obtain 3D object detections in bird's-eye view (BEV). Then, we
introduce hierarchical graph neural networks (GNNs) to track these 3D
detections in BEV for MTMC tracking results. Unlike existing methods, MCBLT has
impressive generalizability across different scenes and diverse camera
settings, with exceptional capability for long-term association handling. As a
result, our proposed MCBLT establishes a new state-of-the-art on the AICity'24
dataset with $81.22$ HOTA, and on the WildTrack dataset with $95.6$ IDF1.
| [
{
"version": "v1",
"created": "Sun, 1 Dec 2024 06:18:06 GMT"
},
{
"version": "v2",
"created": "Sat, 7 Dec 2024 22:46:42 GMT"
},
{
"version": "v3",
"created": "Wed, 26 Mar 2025 19:59:25 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Wang",
"Yizhou",
""
],
[
"Meinhardt",
"Tim",
""
],
[
"Cetintas",
"Orcun",
""
],
[
"Yang",
"Cheng-Yen",
""
],
[
"Pusegaonkar",
"Sameer Satish",
""
],
[
"Missaoui",
"Benjamin",
""
],
[
"Biswas",
"Sujit",
""
],
[
"Tang",
"Zheng",
""
],
[
"Leal-Taixé",
"Laura",
""
]
] | TITLE: MCBLT: Multi-Camera Multi-Object 3D Tracking in Long Videos
ABSTRACT: Object perception from multi-view cameras is crucial for intelligent systems,
particularly in indoor environments, e.g., warehouses, retail stores, and
hospitals. Most traditional multi-target multi-camera (MTMC) detection and
tracking methods rely on 2D object detection, single-view multi-object tracking
(MOT), and cross-view re-identification (ReID) techniques, without properly
handling important 3D information by multi-view image aggregation. In this
paper, we propose a 3D object detection and tracking framework, named MCBLT,
which first aggregates multi-view images with necessary camera calibration
parameters to obtain 3D object detections in bird's-eye view (BEV). Then, we
introduce hierarchical graph neural networks (GNNs) to track these 3D
detections in BEV for MTMC tracking results. Unlike existing methods, MCBLT has
impressive generalizability across different scenes and diverse camera
settings, with exceptional capability for long-term association handling. As a
result, our proposed MCBLT establishes a new state-of-the-art on the AICity'24
dataset with $81.22$ HOTA, and on the WildTrack dataset with $95.6$ IDF1.
|
2412.02479 | Caixin Kang | Caixin Kang, Yubo Chen, Shouwei Ruan, Shiji Zhao, Ruochen Zhang, Jiayi
Wang, Shan Fu, Xingxing Wei | OODFace: Benchmarking Robustness of Face Recognition under Common
Corruptions and Appearance Variations | null | null | null | null | cs.CV cs.AI cs.CR cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With the rise of deep learning, facial recognition technology has seen
extensive research and rapid development. Although facial recognition is
considered a mature technology, we find that existing open-source models and
commercial algorithms lack robustness in certain complex Out-of-Distribution
(OOD) scenarios, raising concerns about the reliability of these systems. In
this paper, we introduce OODFace, which explores the OOD challenges faced by
facial recognition models from two perspectives: common corruptions and
appearance variations. We systematically design 30 OOD scenarios across 9 major
categories tailored for facial recognition. By simulating these challenges on
public datasets, we establish three robustness benchmarks: LFW-C/V, CFP-FP-C/V,
and YTF-C/V. We then conduct extensive experiments on 19 facial recognition
models and 3 commercial APIs, along with extended physical experiments on face
masks to assess their robustness. Next, we explore potential solutions from two
perspectives: defense strategies and Vision-Language Models (VLMs). Based on
the results, we draw several key insights, highlighting the vulnerability of
facial recognition systems to OOD data and suggesting possible solutions.
Additionally, we offer a unified toolkit that includes all corruption and
variation types, easily extendable to other datasets. We hope that our
benchmarks and findings can provide guidance for future improvements in facial
recognition model robustness.
| [
{
"version": "v1",
"created": "Tue, 3 Dec 2024 14:42:31 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 05:40:57 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Kang",
"Caixin",
""
],
[
"Chen",
"Yubo",
""
],
[
"Ruan",
"Shouwei",
""
],
[
"Zhao",
"Shiji",
""
],
[
"Zhang",
"Ruochen",
""
],
[
"Wang",
"Jiayi",
""
],
[
"Fu",
"Shan",
""
],
[
"Wei",
"Xingxing",
""
]
] | TITLE: OODFace: Benchmarking Robustness of Face Recognition under Common
Corruptions and Appearance Variations
ABSTRACT: With the rise of deep learning, facial recognition technology has seen
extensive research and rapid development. Although facial recognition is
considered a mature technology, we find that existing open-source models and
commercial algorithms lack robustness in certain complex Out-of-Distribution
(OOD) scenarios, raising concerns about the reliability of these systems. In
this paper, we introduce OODFace, which explores the OOD challenges faced by
facial recognition models from two perspectives: common corruptions and
appearance variations. We systematically design 30 OOD scenarios across 9 major
categories tailored for facial recognition. By simulating these challenges on
public datasets, we establish three robustness benchmarks: LFW-C/V, CFP-FP-C/V,
and YTF-C/V. We then conduct extensive experiments on 19 facial recognition
models and 3 commercial APIs, along with extended physical experiments on face
masks to assess their robustness. Next, we explore potential solutions from two
perspectives: defense strategies and Vision-Language Models (VLMs). Based on
the results, we draw several key insights, highlighting the vulnerability of
facial recognition systems to OOD data and suggesting possible solutions.
Additionally, we offer a unified toolkit that includes all corruption and
variation types, easily extendable to other datasets. We hope that our
benchmarks and findings can provide guidance for future improvements in facial
recognition model robustness.
|
2412.03044 | Xiaofeng Tan | Xiaofeng Tan, Hongsong Wang, Xin Geng and Liang Wang | Frequency-Guided Diffusion Model with Perturbation Training for
Skeleton-Based Video Anomaly Detection | null | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Video anomaly detection (VAD) is a vital yet complex open-set task in
computer vision, commonly tackled through reconstruction-based methods.
However, these methods struggle with two key limitations: (1) insufficient
robustness in open-set scenarios, where unseen normal motions are frequently
misclassified as anomalies, and (2) an overemphasis on, but restricted capacity
for, local motion reconstruction, which are inherently difficult to capture
accurately due to their diversity. To overcome these challenges, we introduce a
novel frequency-guided diffusion model with perturbation training. First, we
enhance robustness by training a generator to produce perturbed samples, which
are similar to normal samples and target the weakness of the reconstruction
model. This training paradigm expands the reconstruction domain of the model,
improving its generalization to unseen normal motions. Second, to address the
overemphasis on motion details, we employ the 2D Discrete Cosine Transform
(DCT) to separate high-frequency (local) and low-frequency (global) motion
components. By guiding the diffusion model with observed high-frequency
information, we prioritize the reconstruction of low-frequency components,
enabling more accurate and robust anomaly detection. Extensive experiments on
five widely used VAD datasets demonstrate that our approach surpasses
state-of-the-art methods, underscoring its effectiveness in open-set scenarios
and diverse motion contexts. Our project website is
https://xiaofeng-tan.github.io/projects/FG-Diff/index.html.
| [
{
"version": "v1",
"created": "Wed, 4 Dec 2024 05:43:53 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 05:03:14 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Tan",
"Xiaofeng",
""
],
[
"Wang",
"Hongsong",
""
],
[
"Geng",
"Xin",
""
],
[
"Wang",
"Liang",
""
]
] | TITLE: Frequency-Guided Diffusion Model with Perturbation Training for
Skeleton-Based Video Anomaly Detection
ABSTRACT: Video anomaly detection (VAD) is a vital yet complex open-set task in
computer vision, commonly tackled through reconstruction-based methods.
However, these methods struggle with two key limitations: (1) insufficient
robustness in open-set scenarios, where unseen normal motions are frequently
misclassified as anomalies, and (2) an overemphasis on, but restricted capacity
for, local motion reconstruction, which are inherently difficult to capture
accurately due to their diversity. To overcome these challenges, we introduce a
novel frequency-guided diffusion model with perturbation training. First, we
enhance robustness by training a generator to produce perturbed samples, which
are similar to normal samples and target the weakness of the reconstruction
model. This training paradigm expands the reconstruction domain of the model,
improving its generalization to unseen normal motions. Second, to address the
overemphasis on motion details, we employ the 2D Discrete Cosine Transform
(DCT) to separate high-frequency (local) and low-frequency (global) motion
components. By guiding the diffusion model with observed high-frequency
information, we prioritize the reconstruction of low-frequency components,
enabling more accurate and robust anomaly detection. Extensive experiments on
five widely used VAD datasets demonstrate that our approach surpasses
state-of-the-art methods, underscoring its effectiveness in open-set scenarios
and diverse motion contexts. Our project website is
https://xiaofeng-tan.github.io/projects/FG-Diff/index.html.
|
2412.06602 | Tianxin Xie | Tianxin Xie, Yan Rong, Pengfei Zhang, Wenwu Wang, Li Liu | Towards Controllable Speech Synthesis in the Era of Large Language
Models: A Survey | A comprehensive survey on controllable TTS, 26 pages, 7 tables, 6
figures, 317 references. Under review | null | null | null | cs.CL cs.AI cs.LG cs.MM cs.SD eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Text-to-speech (TTS), also known as speech synthesis, is a prominent research
area that aims to generate natural-sounding human speech from text. Recently,
with the increasing industrial demand, TTS technologies have evolved beyond
synthesizing human-like speech to enabling controllable speech generation. This
includes fine-grained control over various attributes of synthesized speech
such as emotion, prosody, timbre, and duration. In addition, advancements in
deep learning, such as diffusion and large language models, have significantly
enhanced controllable TTS over the past several years. In this work, we conduct
a comprehensive survey of controllable TTS, covering approaches ranging from
basic control techniques to methods utilizing natural language prompts, aiming
to provide a clear understanding of the current state of research. We examine
the general controllable TTS pipeline, challenges, model architectures, and
control strategies, offering a comprehensive and clear taxonomy of existing
methods. Additionally, we provide a detailed summary of datasets and evaluation
metrics and shed some light on the applications and future directions of
controllable TTS. To the best of our knowledge, this survey paper provides the
first comprehensive review of emerging controllable TTS methods, which can
serve as a beneficial resource for both academic researchers and industrial
practitioners.
| [
{
"version": "v1",
"created": "Mon, 9 Dec 2024 15:50:25 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 03:56:00 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Xie",
"Tianxin",
""
],
[
"Rong",
"Yan",
""
],
[
"Zhang",
"Pengfei",
""
],
[
"Wang",
"Wenwu",
""
],
[
"Liu",
"Li",
""
]
] | TITLE: Towards Controllable Speech Synthesis in the Era of Large Language
Models: A Survey
ABSTRACT: Text-to-speech (TTS), also known as speech synthesis, is a prominent research
area that aims to generate natural-sounding human speech from text. Recently,
with the increasing industrial demand, TTS technologies have evolved beyond
synthesizing human-like speech to enabling controllable speech generation. This
includes fine-grained control over various attributes of synthesized speech
such as emotion, prosody, timbre, and duration. In addition, advancements in
deep learning, such as diffusion and large language models, have significantly
enhanced controllable TTS over the past several years. In this work, we conduct
a comprehensive survey of controllable TTS, covering approaches ranging from
basic control techniques to methods utilizing natural language prompts, aiming
to provide a clear understanding of the current state of research. We examine
the general controllable TTS pipeline, challenges, model architectures, and
control strategies, offering a comprehensive and clear taxonomy of existing
methods. Additionally, we provide a detailed summary of datasets and evaluation
metrics and shed some light on the applications and future directions of
controllable TTS. To the best of our knowledge, this survey paper provides the
first comprehensive review of emerging controllable TTS methods, which can
serve as a beneficial resource for both academic researchers and industrial
practitioners.
|
2412.09599 | Ayaka Higami | Ayaka Higami, Karin Oshima, Tomoyo Isoguchi Shiramatsu, Hirokazu
Takahashi, Shohei Nobuhara, Ko Nishino | RatBodyFormer: Rat Body Surface from Keypoints | https://vision.ist.i.kyoto-u.ac.jp/research/ratbodyformer/ | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Analyzing rat behavior lies at the heart of many scientific studies. Past
methods for automated rodent modeling have focused on 3D pose estimation from
keypoints, e.g., face and appendages. The pose, however, does not capture the
rich body surface movement encoding the subtle rat behaviors like curling and
stretching. The body surface lacks features that can be visually defined,
evading these established keypoint-based methods. In this paper, we introduce
the first method for reconstructing the rat body surface as a dense set of
points by learning to predict it from the sparse keypoints that can be detected
with past methods. Our method consists of two key contributions. The first is
RatDome, a novel multi-camera system for rat behavior capture, and a
large-scale dataset captured with it that consists of pairs of 3D keypoints and
3D body surface points. The second is RatBodyFormer, a novel network to
transform detected keypoints to 3D body surface points. RatBodyFormer is
agnostic to the exact locations of the 3D body surface points in the training
data and is trained with masked-learning. We experimentally validate our
framework with a number of real-world experiments. Our results collectively
serve as a novel foundation for automated rat behavior analysis.
| [
{
"version": "v1",
"created": "Thu, 12 Dec 2024 18:59:00 GMT"
},
{
"version": "v2",
"created": "Wed, 18 Dec 2024 03:49:22 GMT"
},
{
"version": "v3",
"created": "Thu, 27 Mar 2025 01:58:34 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Higami",
"Ayaka",
""
],
[
"Oshima",
"Karin",
""
],
[
"Shiramatsu",
"Tomoyo Isoguchi",
""
],
[
"Takahashi",
"Hirokazu",
""
],
[
"Nobuhara",
"Shohei",
""
],
[
"Nishino",
"Ko",
""
]
] | TITLE: RatBodyFormer: Rat Body Surface from Keypoints
ABSTRACT: Analyzing rat behavior lies at the heart of many scientific studies. Past
methods for automated rodent modeling have focused on 3D pose estimation from
keypoints, e.g., face and appendages. The pose, however, does not capture the
rich body surface movement encoding the subtle rat behaviors like curling and
stretching. The body surface lacks features that can be visually defined,
evading these established keypoint-based methods. In this paper, we introduce
the first method for reconstructing the rat body surface as a dense set of
points by learning to predict it from the sparse keypoints that can be detected
with past methods. Our method consists of two key contributions. The first is
RatDome, a novel multi-camera system for rat behavior capture, and a
large-scale dataset captured with it that consists of pairs of 3D keypoints and
3D body surface points. The second is RatBodyFormer, a novel network to
transform detected keypoints to 3D body surface points. RatBodyFormer is
agnostic to the exact locations of the 3D body surface points in the training
data and is trained with masked-learning. We experimentally validate our
framework with a number of real-world experiments. Our results collectively
serve as a novel foundation for automated rat behavior analysis.
|
2412.15215 | Tao Xie | Tao Xie, Xi Chen, Zhen Xu, Yiman Xie, Yudong Jin, Yujun Shen, Sida
Peng, Hujun Bao, Xiaowei Zhou | EnvGS: Modeling View-Dependent Appearance with Environment Gaussian | Project page: https://zju3dv.github.io/envgs | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Reconstructing complex reflections in real-world scenes from 2D images is
essential for achieving photorealistic novel view synthesis. Existing methods
that utilize environment maps to model reflections from distant lighting often
struggle with high-frequency reflection details and fail to account for
near-field reflections. In this work, we introduce EnvGS, a novel approach that
employs a set of Gaussian primitives as an explicit 3D representation for
capturing reflections of environments. These environment Gaussian primitives
are incorporated with base Gaussian primitives to model the appearance of the
whole scene. To efficiently render these environment Gaussian primitives, we
developed a ray-tracing-based renderer that leverages the GPU's RT core for
fast rendering. This allows us to jointly optimize our model for high-quality
reconstruction while maintaining real-time rendering speeds. Results from
multiple real-world and synthetic datasets demonstrate that our method produces
significantly more detailed reflections, achieving the best rendering quality
in real-time novel view synthesis. The code is available at
https://zju3dv.github.io/envgs.
| [
{
"version": "v1",
"created": "Thu, 19 Dec 2024 18:59:57 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 11:12:07 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Xie",
"Tao",
""
],
[
"Chen",
"Xi",
""
],
[
"Xu",
"Zhen",
""
],
[
"Xie",
"Yiman",
""
],
[
"Jin",
"Yudong",
""
],
[
"Shen",
"Yujun",
""
],
[
"Peng",
"Sida",
""
],
[
"Bao",
"Hujun",
""
],
[
"Zhou",
"Xiaowei",
""
]
] | TITLE: EnvGS: Modeling View-Dependent Appearance with Environment Gaussian
ABSTRACT: Reconstructing complex reflections in real-world scenes from 2D images is
essential for achieving photorealistic novel view synthesis. Existing methods
that utilize environment maps to model reflections from distant lighting often
struggle with high-frequency reflection details and fail to account for
near-field reflections. In this work, we introduce EnvGS, a novel approach that
employs a set of Gaussian primitives as an explicit 3D representation for
capturing reflections of environments. These environment Gaussian primitives
are incorporated with base Gaussian primitives to model the appearance of the
whole scene. To efficiently render these environment Gaussian primitives, we
developed a ray-tracing-based renderer that leverages the GPU's RT core for
fast rendering. This allows us to jointly optimize our model for high-quality
reconstruction while maintaining real-time rendering speeds. Results from
multiple real-world and synthetic datasets demonstrate that our method produces
significantly more detailed reflections, achieving the best rendering quality
in real-time novel view synthesis. The code is available at
https://zju3dv.github.io/envgs.
|
2412.16218 | Xinkai Wei | Jianqing Liang, Xinkai Wei, Min Chen, Zhiqiang Wang, Jiye Liang | GNN-Transformer Cooperative Architecture for Trustworthy Graph
Contrastive Learning | In Proceedings of AAAI 2025 | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Graph contrastive learning (GCL) has become a hot topic in the field of graph
representation learning. In contrast to traditional supervised learning relying
on a large number of labels, GCL exploits augmentation strategies to generate
multiple views and positive/negative pairs, both of which greatly influence the
performance. Unfortunately, commonly used random augmentations may disturb the
underlying semantics of graphs. Moreover, traditional GNNs, a type of widely
employed encoders in GCL, are inevitably confronted with over-smoothing and
over-squashing problems. To address these issues, we propose GNN-Transformer
Cooperative Architecture for Trustworthy Graph Contrastive Learning (GTCA),
which inherits the advantages of both GNN and Transformer, incorporating graph
topology to obtain comprehensive graph representations. Theoretical analysis
verifies the trustworthiness of the proposed method. Extensive experiments on
benchmark datasets demonstrate state-of-the-art empirical performance.
| [
{
"version": "v1",
"created": "Wed, 18 Dec 2024 09:20:12 GMT"
},
{
"version": "v2",
"created": "Tue, 24 Dec 2024 02:02:24 GMT"
},
{
"version": "v3",
"created": "Tue, 28 Jan 2025 09:48:54 GMT"
},
{
"version": "v4",
"created": "Thu, 27 Mar 2025 13:44:56 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Liang",
"Jianqing",
""
],
[
"Wei",
"Xinkai",
""
],
[
"Chen",
"Min",
""
],
[
"Wang",
"Zhiqiang",
""
],
[
"Liang",
"Jiye",
""
]
] | TITLE: GNN-Transformer Cooperative Architecture for Trustworthy Graph
Contrastive Learning
ABSTRACT: Graph contrastive learning (GCL) has become a hot topic in the field of graph
representation learning. In contrast to traditional supervised learning relying
on a large number of labels, GCL exploits augmentation strategies to generate
multiple views and positive/negative pairs, both of which greatly influence the
performance. Unfortunately, commonly used random augmentations may disturb the
underlying semantics of graphs. Moreover, traditional GNNs, a type of widely
employed encoders in GCL, are inevitably confronted with over-smoothing and
over-squashing problems. To address these issues, we propose GNN-Transformer
Cooperative Architecture for Trustworthy Graph Contrastive Learning (GTCA),
which inherits the advantages of both GNN and Transformer, incorporating graph
topology to obtain comprehensive graph representations. Theoretical analysis
verifies the trustworthiness of the proposed method. Extensive experiments on
benchmark datasets demonstrate state-of-the-art empirical performance.
|
2412.20104 | Wenkun He | Wenkun He, Yun Liu, Ruitao Liu, Li Yi | SyncDiff: Synchronized Motion Diffusion for Multi-Body Human-Object
Interaction Synthesis | 26 pages, 10 figures | null | null | null | cs.CV cs.AI cs.LG cs.RO | http://creativecommons.org/licenses/by/4.0/ | Synthesizing realistic human-object interaction motions is a critical problem
in VR/AR and human animation. Unlike the commonly studied scenarios involving a
single human or hand interacting with one object, we address a more generic
multi-body setting with arbitrary numbers of humans, hands, and objects. This
complexity introduces significant challenges in synchronizing motions due to
the high correlations and mutual influences among bodies. To address these
challenges, we introduce SyncDiff, a novel method for multi-body interaction
synthesis using a synchronized motion diffusion strategy. SyncDiff employs a
single diffusion model to capture the joint distribution of multi-body motions.
To enhance motion fidelity, we propose a frequency-domain motion decomposition
scheme. Additionally, we introduce a new set of alignment scores to emphasize
the synchronization of different body motions. SyncDiff jointly optimizes both
data sample likelihood and alignment likelihood through an explicit
synchronization strategy. Extensive experiments across four datasets with
various multi-body configurations demonstrate the superiority of SyncDiff over
existing state-of-the-art motion synthesis methods.
| [
{
"version": "v1",
"created": "Sat, 28 Dec 2024 10:12:12 GMT"
},
{
"version": "v2",
"created": "Mon, 13 Jan 2025 11:46:06 GMT"
},
{
"version": "v3",
"created": "Tue, 25 Mar 2025 04:15:15 GMT"
},
{
"version": "v4",
"created": "Thu, 27 Mar 2025 02:17:08 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"He",
"Wenkun",
""
],
[
"Liu",
"Yun",
""
],
[
"Liu",
"Ruitao",
""
],
[
"Yi",
"Li",
""
]
] | TITLE: SyncDiff: Synchronized Motion Diffusion for Multi-Body Human-Object
Interaction Synthesis
ABSTRACT: Synthesizing realistic human-object interaction motions is a critical problem
in VR/AR and human animation. Unlike the commonly studied scenarios involving a
single human or hand interacting with one object, we address a more generic
multi-body setting with arbitrary numbers of humans, hands, and objects. This
complexity introduces significant challenges in synchronizing motions due to
the high correlations and mutual influences among bodies. To address these
challenges, we introduce SyncDiff, a novel method for multi-body interaction
synthesis using a synchronized motion diffusion strategy. SyncDiff employs a
single diffusion model to capture the joint distribution of multi-body motions.
To enhance motion fidelity, we propose a frequency-domain motion decomposition
scheme. Additionally, we introduce a new set of alignment scores to emphasize
the synchronization of different body motions. SyncDiff jointly optimizes both
data sample likelihood and alignment likelihood through an explicit
synchronization strategy. Extensive experiments across four datasets with
various multi-body configurations demonstrate the superiority of SyncDiff over
existing state-of-the-art motion synthesis methods.
|
2501.01855 | Huaxiang Zhang | Huaxiang Zhang, Kai Liu, Zhongxue Gan, and Guo-Niu Zhu | UAV-DETR: Efficient End-to-End Object Detection for Unmanned Aerial
Vehicle Imagery | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Unmanned aerial vehicle object detection (UAV-OD) has been widely used in
various scenarios. However, most existing UAV-OD algorithms rely on manually
designed components, which require extensive tuning. End-to-end models that do
not depend on such manually designed components are mainly designed for natural
images, which are less effective for UAV imagery. To address such challenges,
this paper proposes an efficient detection transformer (DETR) framework
tailored for UAV imagery, i.e., UAV-DETR. The framework includes a multi-scale
feature fusion with frequency enhancement module, which captures both spatial
and frequency information at different scales. In addition, a frequency-focused
down-sampling module is presented to retain critical spatial details during
down-sampling. A semantic alignment and calibration module is developed to
align and fuse features from different fusion paths. Experimental results
demonstrate the effectiveness and generalization of our approach across various
UAV imagery datasets. On the VisDrone dataset, our method improves AP by 3.1\%
and $\text{AP}_{50}$ by 4.2\% over the baseline. Similar enhancements are
observed on the UAVVaste dataset. The project page:
https://github.com/ValiantDiligent/UAV-DETR
| [
{
"version": "v1",
"created": "Fri, 3 Jan 2025 15:11:14 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 14:17:42 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Zhang",
"Huaxiang",
""
],
[
"Liu",
"Kai",
""
],
[
"Gan",
"Zhongxue",
""
],
[
"Zhu",
"Guo-Niu",
""
]
] | TITLE: UAV-DETR: Efficient End-to-End Object Detection for Unmanned Aerial
Vehicle Imagery
ABSTRACT: Unmanned aerial vehicle object detection (UAV-OD) has been widely used in
various scenarios. However, most existing UAV-OD algorithms rely on manually
designed components, which require extensive tuning. End-to-end models that do
not depend on such manually designed components are mainly designed for natural
images, which are less effective for UAV imagery. To address such challenges,
this paper proposes an efficient detection transformer (DETR) framework
tailored for UAV imagery, i.e., UAV-DETR. The framework includes a multi-scale
feature fusion with frequency enhancement module, which captures both spatial
and frequency information at different scales. In addition, a frequency-focused
down-sampling module is presented to retain critical spatial details during
down-sampling. A semantic alignment and calibration module is developed to
align and fuse features from different fusion paths. Experimental results
demonstrate the effectiveness and generalization of our approach across various
UAV imagery datasets. On the VisDrone dataset, our method improves AP by 3.1\%
and $\text{AP}_{50}$ by 4.2\% over the baseline. Similar enhancements are
observed on the UAVVaste dataset. The project page:
https://github.com/ValiantDiligent/UAV-DETR
|
2501.02471 | Yishen Liu | Yishen Liu and Shengda Luo and Zishao Zhong and Tongtong Wu and
Jianguo Zhang and Peiyao Ou and Yong Liang and Liang Liu and Hudan Pan | Hengqin-RA-v1: Advanced Large Language Model for Diagnosis and Treatment
of Rheumatoid Arthritis with Dataset based Traditional Chinese Medicine | 8 pages, 5 figures, AAAI-2025 Workshop | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | Large language models (LLMs) primarily trained on English texts, often face
biases and inaccuracies in Chinese contexts. Their limitations are pronounced
in fields like Traditional Chinese Medicine (TCM), where cultural and clinical
subtleties are vital, further hindered by a lack of domain-specific data, such
as rheumatoid arthritis (RA). To address these issues, this paper introduces
Hengqin-RA-v1, the first large language model specifically tailored for TCM
with a focus on diagnosing and treating RA. We also present HQ-GCM-RA-C1, a
comprehensive RA-specific dataset curated from ancient Chinese medical
literature, classical texts, and modern clinical studies. This dataset empowers
Hengqin-RA-v1 to deliver accurate and culturally informed responses,
effectively bridging the gaps left by general-purpose models. Extensive
experiments demonstrate that Hengqin-RA-v1 outperforms state-of-the-art models,
even surpassing the diagnostic accuracy of TCM practitioners in certain cases.
| [
{
"version": "v1",
"created": "Sun, 5 Jan 2025 07:46:51 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 06:39:45 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Liu",
"Yishen",
""
],
[
"Luo",
"Shengda",
""
],
[
"Zhong",
"Zishao",
""
],
[
"Wu",
"Tongtong",
""
],
[
"Zhang",
"Jianguo",
""
],
[
"Ou",
"Peiyao",
""
],
[
"Liang",
"Yong",
""
],
[
"Liu",
"Liang",
""
],
[
"Pan",
"Hudan",
""
]
] | TITLE: Hengqin-RA-v1: Advanced Large Language Model for Diagnosis and Treatment
of Rheumatoid Arthritis with Dataset based Traditional Chinese Medicine
ABSTRACT: Large language models (LLMs) primarily trained on English texts, often face
biases and inaccuracies in Chinese contexts. Their limitations are pronounced
in fields like Traditional Chinese Medicine (TCM), where cultural and clinical
subtleties are vital, further hindered by a lack of domain-specific data, such
as rheumatoid arthritis (RA). To address these issues, this paper introduces
Hengqin-RA-v1, the first large language model specifically tailored for TCM
with a focus on diagnosing and treating RA. We also present HQ-GCM-RA-C1, a
comprehensive RA-specific dataset curated from ancient Chinese medical
literature, classical texts, and modern clinical studies. This dataset empowers
Hengqin-RA-v1 to deliver accurate and culturally informed responses,
effectively bridging the gaps left by general-purpose models. Extensive
experiments demonstrate that Hengqin-RA-v1 outperforms state-of-the-art models,
even surpassing the diagnostic accuracy of TCM practitioners in certain cases.
|
2501.03550 | Sha Wang | Pan Guo, Yuan Gao, Yongjie Pu, Zhigang Zhao, Zhenhua Cong and Sha Wang | Intelligent Mode-Locked Single-Cavity Dual-Comb Laser Utilizing
Time-Stretch Dispersive Fourier Transform Spectroscopy with Supplemental File | 10 pages, 8 figures | null | null | null | physics.optics | http://creativecommons.org/licenses/by/4.0/ | As dual combs play a significant role in numerous high-precision
measurements, their efficient generation has been widely researched. Although
the single-cavity dual-comb generation can avoid the complex active
stabilization methods, achieving and maintaining stable dual-comb mode locking
within a single cavity remains a critical challenge. To break through this
constraint, a two-part evaluation criterion containing a fitness function and a
CNN-Transformer network is employed to achieve mode locking and classify the
dual-comb mode-locked state. Simulated time-stretch dispersive Fourier
transform (DFT) spectra are used as datasets, which simplifies the optimization
process and does not rely on specific experimental data. A developed
evolutionary algorithm (EA) for paddle-based motorized polarization controllers
(MPCs) is proposed, enabling the intelligent attainment of dual-comb
mode-locked states. A real-time library stores fitness and MPC angles,
facilitating mode-locked state achievement within 2 seconds. Finally, long term
running of dual-comb mode locking is ensured by a random collision algorithm
utilizing an evaluation criterion of weak soliton peaks.
| [
{
"version": "v1",
"created": "Tue, 7 Jan 2025 05:52:35 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 12:35:53 GMT"
}
] | 2025-03-28T00:00:00 | [
[
"Guo",
"Pan",
""
],
[
"Gao",
"Yuan",
""
],
[
"Pu",
"Yongjie",
""
],
[
"Zhao",
"Zhigang",
""
],
[
"Cong",
"Zhenhua",
""
],
[
"Wang",
"Sha",
""
]
] | TITLE: Intelligent Mode-Locked Single-Cavity Dual-Comb Laser Utilizing
Time-Stretch Dispersive Fourier Transform Spectroscopy with Supplemental File
ABSTRACT: As dual combs play a significant role in numerous high-precision
measurements, their efficient generation has been widely researched. Although
the single-cavity dual-comb generation can avoid the complex active
stabilization methods, achieving and maintaining stable dual-comb mode locking
within a single cavity remains a critical challenge. To break through this
constraint, a two-part evaluation criterion containing a fitness function and a
CNN-Transformer network is employed to achieve mode locking and classify the
dual-comb mode-locked state. Simulated time-stretch dispersive Fourier
transform (DFT) spectra are used as datasets, which simplifies the optimization
process and does not rely on specific experimental data. A developed
evolutionary algorithm (EA) for paddle-based motorized polarization controllers
(MPCs) is proposed, enabling the intelligent attainment of dual-comb
mode-locked states. A real-time library stores fitness and MPC angles,
facilitating mode-locked state achievement within 2 seconds. Finally, long term
running of dual-comb mode locking is ensured by a random collision algorithm
utilizing an evaluation criterion of weak soliton peaks.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.