Datasets:

Languages:
English
ArXiv:
License:
ziyjiang commited on
Commit
2dc3f6e
Β·
verified Β·
1 Parent(s): 47b185c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +41 -8
README.md CHANGED
@@ -8,13 +8,16 @@ language:
8
 
9
  # MMEB-V2 (Massive Multimodal Embedding Benchmark)
10
 
11
- Building upon on our original [MMEB](https://arxiv.org/abs/2410.05160), MMEB-V2 expands the evaluation scope to include five new tasks: four video-based tasks β€” Video Retrieval, Moment Retrieval, Video Classification, and Video Question Answering β€” and one task focused on visual documents, Visual Document Retrieval. This comprehensive suite enables robust evaluation of multimodal embedding models across static, temporal, and structured visual data settings.
 
 
 
12
 
13
  |[**Github**](https://github.com/TIGER-AI-Lab/VLM2Vec) | [**πŸ†Leaderboard**](https://huggingface.co/spaces/TIGER-Lab/MMEB) | [**πŸ“–MMEB-V2/VLM2Vec-V2 Paper (TBA)**](https://arxiv.org/abs/2410.05160) | | [**πŸ“–MMEB-V1/VLM2Vec-V1 Paper**](https://arxiv.org/abs/2410.05160) |
14
 
15
 
16
  ## πŸš€ What's New
17
- - **\[2025.05\]** Initial release.
18
 
19
 
20
  ## Dataset Overview
@@ -25,7 +28,10 @@ We present an overview of the MMEB-V2 dataset below:
25
 
26
  ## Dataset Structure
27
 
28
- Below is the directory structure for this Huggingface Repo:
 
 
 
29
  ```
30
 
31
  β†’ video-tasks/
@@ -34,15 +40,42 @@ Below is the directory structure for this Huggingface Repo:
34
  β”‚ β”œβ”€β”€ video_qa.tar.gz
35
  β”‚ β”œβ”€β”€ video_ret.tar.gz
36
  β”‚ └── video_mret.tar.gz
37
- β”œβ”€β”€ raw videos/ (TBA)
38
- β”‚ β”œβ”€β”€ video_cls.tar.gz
39
- β”‚ β”œβ”€β”€ video_qa.tar.gz
40
- β”‚ β”œβ”€β”€ video_ret.tar.gz
41
- β”‚ └── video_mret.tar.gz
42
 
43
  β†’ image-tasks/
44
  β”œβ”€β”€ mmeb_v1.tar.gz
45
  └── visdoc.tar.gz
46
 
 
 
 
 
47
 
48
  ```
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8
 
9
  # MMEB-V2 (Massive Multimodal Embedding Benchmark)
10
 
11
+ Building upon on our original [**MMEB**](https://arxiv.org/abs/2410.05160), **MMEB-V2** expands the evaluation scope to include five new tasks: four video-based tasks β€” Video Retrieval, Moment Retrieval, Video Classification, and Video Question Answering β€” and one task focused on visual documents, Visual Document Retrieval. This comprehensive suite enables robust evaluation of multimodal embedding models across static, temporal, and structured visual data settings.
12
+
13
+ **This Hugging Face repository contains only the raw image and video files used in MMEB-V2, which need to be downloaded in advance.**
14
+ The test data for each task in MMEB-V2 is available [here](https://huggingface.co/VLM2Vec) and will be automatically downloaded and used by our code. More details on how to set it up are provided in the following sections.
15
 
16
  |[**Github**](https://github.com/TIGER-AI-Lab/VLM2Vec) | [**πŸ†Leaderboard**](https://huggingface.co/spaces/TIGER-Lab/MMEB) | [**πŸ“–MMEB-V2/VLM2Vec-V2 Paper (TBA)**](https://arxiv.org/abs/2410.05160) | | [**πŸ“–MMEB-V1/VLM2Vec-V1 Paper**](https://arxiv.org/abs/2410.05160) |
17
 
18
 
19
  ## πŸš€ What's New
20
+ - **\[2025.05\]** Initial release of MMEB-V2.
21
 
22
 
23
  ## Dataset Overview
 
28
 
29
  ## Dataset Structure
30
 
31
+ The directory structure of this Hugging Face repository is shown below.
32
+ For video tasks, we provide both sampled frames and raw videos (the latter will be released later). For image tasks, we provide the raw images.
33
+ Files from each meta-task are zipped together, resulting in six files. For example, ``video_cls.tar.gz`` contains the sampled frames for the video classification task.
34
+
35
  ```
36
 
37
  β†’ video-tasks/
 
40
  β”‚ β”œβ”€β”€ video_qa.tar.gz
41
  β”‚ β”œβ”€β”€ video_ret.tar.gz
42
  β”‚ └── video_mret.tar.gz
43
+ β”œβ”€β”€ raw videos/ (To be released)
 
 
 
 
44
 
45
  β†’ image-tasks/
46
  β”œβ”€β”€ mmeb_v1.tar.gz
47
  └── visdoc.tar.gz
48
 
49
+ ```
50
+
51
+ After downloading and unzipping these files locally, you can organize them as shown below. (You may choose to use ``Git LFS`` or ``wget`` for downloading.)
52
+ Then, simply specify the correct file path in the configuration file used by your code.
53
 
54
  ```
55
+
56
+ β†’ MMEB
57
+ β”œβ”€β”€ video-tasks/
58
+ β”‚ └── frames/
59
+ β”‚ β”œβ”€β”€ video_cls/
60
+ β”‚ β”‚ β”œβ”€β”€ UCF101/
61
+ β”‚ β”‚ β”‚ └── video_1/ # video ID
62
+ β”‚ β”‚ β”‚ β”œβ”€β”€ frame1.png # frame from video_1
63
+ β”‚ β”‚ β”‚ β”œβ”€β”€ frame2.png
64
+ β”‚ β”‚ β”‚ └── ...
65
+ β”‚ β”‚ β”œβ”€β”€ HMDB51/
66
+ β”‚ β”‚ β”œβ”€β”€ Breakfast/
67
+ β”‚ β”‚ └── ... # other datasets from video classification category
68
+ β”‚ β”œβ”€β”€ video_qa/
69
+ β”‚ β”‚ └── ... # video QA datasets
70
+ β”‚ β”œβ”€β”€ video_ret/
71
+ β”‚ β”‚ └── ... # video retrieval datasets
72
+ β”‚ └── video_mret/
73
+ β”‚ └── ... # moment retrieval datasets
74
+ β”œβ”€β”€ image-tasks/
75
+ β”‚ β”œβ”€β”€ mmeb_v1/
76
+ β”‚ β”‚ └── ... # MMEB-V1 datasets
77
+ β”‚ └── visdoc/
78
+ β”‚ └── ... # visual document retrieval datasets
79
+
80
+
81
+ ```