Datasets:
Update README.md
Browse files
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) | [**πLeaderboard**](https://huggingface.co/spaces/TIGER-Lab/MMEB) | [**πMMEB-V2/VLM2Vec-V2 Paper (TBA)**](https://arxiv.org/abs/2410.05160) | | [**πMMEB-V1/VLM2Vec-V1 Paper**](https://arxiv.org/abs/2410.05160) |
|
14 |
|
15 |
|
16 |
## π 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 |
-
|
|
|
|
|
|
|
29 |
```
|
30 |
|
31 |
β video-tasks/
|
@@ -34,15 +40,42 @@ Below is the directory structure for this Huggingface Repo:
|
|
34 |
β βββ video_qa.tar.gz
|
35 |
β βββ video_ret.tar.gz
|
36 |
β βββ video_mret.tar.gz
|
37 |
-
βββ raw videos/ (
|
38 |
-
β βββ video_cls.tar.gz
|
39 |
-
β βββ video_qa.tar.gz
|
40 |
-
β βββ video_ret.tar.gz
|
41 |
-
β βββ video_mret.tar.gz
|
42 |
|
43 |
β image-tasks/
|
44 |
βββ mmeb_v1.tar.gz
|
45 |
βββ 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) | [**πLeaderboard**](https://huggingface.co/spaces/TIGER-Lab/MMEB) | [**πMMEB-V2/VLM2Vec-V2 Paper (TBA)**](https://arxiv.org/abs/2410.05160) | | [**πMMEB-V1/VLM2Vec-V1 Paper**](https://arxiv.org/abs/2410.05160) |
|
17 |
|
18 |
|
19 |
## π 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 βββ video_qa.tar.gz
|
41 |
β βββ video_ret.tar.gz
|
42 |
β βββ video_mret.tar.gz
|
43 |
+
βββ raw videos/ (To be released)
|
|
|
|
|
|
|
|
|
44 |
|
45 |
β image-tasks/
|
46 |
βββ mmeb_v1.tar.gz
|
47 |
βββ 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 |
+
βββ video-tasks/
|
58 |
+
β βββ frames/
|
59 |
+
β βββ video_cls/
|
60 |
+
β β βββ UCF101/
|
61 |
+
β β β βββ video_1/ # video ID
|
62 |
+
β β β βββ frame1.png # frame from video_1
|
63 |
+
β β β βββ frame2.png
|
64 |
+
β β β βββ ...
|
65 |
+
β β βββ HMDB51/
|
66 |
+
β β βββ Breakfast/
|
67 |
+
β β βββ ... # other datasets from video classification category
|
68 |
+
β βββ video_qa/
|
69 |
+
β β βββ ... # video QA datasets
|
70 |
+
β βββ video_ret/
|
71 |
+
β β βββ ... # video retrieval datasets
|
72 |
+
β βββ video_mret/
|
73 |
+
β βββ ... # moment retrieval datasets
|
74 |
+
βββ image-tasks/
|
75 |
+
β βββ mmeb_v1/
|
76 |
+
β β βββ ... # MMEB-V1 datasets
|
77 |
+
β βββ visdoc/
|
78 |
+
β βββ ... # visual document retrieval datasets
|
79 |
+
|
80 |
+
|
81 |
+
```
|