add first_frame_mask

.gitattributes

@@ -57,3 +57,9 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+*/*/*.json filter=lfs diff=lfs merge=lfs -text
+*/*.json filter=lfs diff=lfs merge=lfs -text
+assets/statistics_2.png filter=lfs diff=lfs merge=lfs -text
+assets/wan-move-logo.png filter=lfs diff=lfs merge=lfs -text
+assets/construction.png filter=lfs diff=lfs merge=lfs -text
+assets/statistics_1.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -23,11 +23,11 @@ tags:
 ## MoveBench: A Comprehensive and Well-Curated Benchmark to Access Motion Control in Videos
 
 
-MoveBench evaluates fine-grained point-level motion control in generated videos. We categorize the video library from [Pexels](https://www.pexels.com/videos/) into 54 content categories, 10-25 videos each, giving rise to 1018 cases to ensure a broad scenario coverage. All video clips maintain a 5-second duration to facilitate evaluation of long-range dynamics. Every clip is paired with detailed motion annotations for a single object. Addtional 192 clips have motion annotations for multiple objects. We ensure annotation quality by developing an interactive labeling pipeline, marrying annotation precision with automated scalability.
-
- Welcome everyone to use it!
+MoveBench evaluates fine-grained point-level motion control in generated videos. We categorize the video library from [Pexels](https://www.pexels.com/videos/) into 54 content categories, 10-25 videos each, giving rise to 1018 cases to ensure a broad scenario coverage. All video clips maintain a 5-second duration to facilitate evaluation of long-range dynamics. Every clip is paired with detailed motion annotations for a single object. Addtional 192 clips have motion annotations for multiple objects.
 
+To annotate, annotators click on a target region in the first frame, prompting SAM to generate an initial mask. When the mask exceeds the desired area, annotators add negative points to exclude irrelevant regions. After annotation, each video contains at least one point trajectory indicating a representative motion, with 192 videos additionally including multiple-object motion trajectories.
 
+Welcome everyone to use it!
 
 
 ## Statistics
@@ -72,6 +72,15 @@ MoveBench
 │   │   ├── Pexels_videoid_0.jpg
 │   │   ├── Pexels_videoid_1.jpg
 │   │   ├── ...
+│   ├── first_frame_mask
+│   │   ├── single
+│   │   │   ├── Pexels_videoid_0_mask_idx.png
+│   │   │   ├── Pexels_videoid_1_mask_idx.png
+│   │   │   ├── ...
+│   │   ├── multi
+│   │   │   ├── Pexels_videoid_0_mask_idx.png
+│   │   │   ├── Pexels_videoid_1_mask_idx.png
+│   │   │   ├── ...
 │   ├── video
 │   │   ├── Pexels_videoid_0.mp4
 │   │   ├── Pexels_videoid_1.mp4
@@ -92,6 +101,15 @@ MoveBench
 │   │   ├── Pexels_videoid_0.jpg
 │   │   ├── Pexels_videoid_1.jpg
 │   │   ├── ...
+│   ├── first_frame_mask
+│   │   ├── single
+│   │   │   ├── Pexels_videoid_0_mask_idx.png
+│   │   │   ├── Pexels_videoid_1_mask_idx.png
+│   │   │   ├── ...
+│   │   ├── multi
+│   │   │   ├── Pexels_videoid_0_mask_idx.png
+│   │   │   ├── Pexels_videoid_1_mask_idx.png
+│   │   │   ├── ...
 │   ├── video
 │   │   ├── Pexels_videoid_0.mp4
 │   │   ├── Pexels_videoid_1.mp4

bench.py

@@ -1,3 +1,4 @@
+huggingface-cli lfs-enable-largefiles .
 import os
 os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'
 import re

en.tar.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6d0196c7c5c51fd83b777e57db679ed340d11ba46bf3892e8f38dcd33a0eef67
-size 666265297
+oid sha256:85b036492653970bd37c76baea049940e917bb9845866d157c0db8d0d5d04875
+size 668580196

zh.tar.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:65989a137e1f7f227f2cd0161288529d00521663077706aac521498ca2bec49c
-size 666270320
+oid sha256:0351a877756caff2f831d27115466ff3fc2bd7a5bec0d06a35490ce6538cc0a7
+size 668587315