Hugging Face's logo

Datasets:
ml-resources
/
SACo-VEval

Modalities:
Text
Languages:
English
Size:
< 1K
License:
Dataset card Data Studio Files
xet
 Community
Dataset Viewer
Auto-converted to Parquet Duplicate
info
dict
videos
list
annotations
list
categories
list
video_np_pairs
list
{ "version": "v1", "date": "2025-09-24T00:00:00", "description": "SA-Co/VEval SA-V Test" }
[{"id":0,"video_name":"sav_000208","file_names":["sav_000208/00000.jpg","sav_000208/00004.jpg","sav_(...TRUNCATED)
[{"id":0,"segmentations":[{"counts":"`^`>b0[o04M4M10O100O100O1O2O0O101N101N3M3Kbam2","size":[1024,57(...TRUNCATED)
[{"id":1,"name":"a monkey"},{"id":2,"name":"a horse"},{"id":3,"name":"a fan"},{"id":4,"name":"a pile(...TRUNCATED)
[{"id":0,"video_id":35,"category_id":50912,"noun_phrase":"an image of bottle","num_masklets":0},{"id(...TRUNCATED)

Mirror of facebook/SACo-VEval. All credit to the original authors. Not an official release from them. License unchanged (SAM License); see LICENSE.

SA-Co/VEval Dataset

License each domain has its own License

  • SA-Co/VEval - SA-V: CC-BY-NC 4.0
  • SA-Co/VEval - YT-Temporal-1B: CC-BY-NC 4.0
  • SA-Co/VEval - SmartGlasses: CC-by-4.0

SA-Co/VEval is an evaluation dataset comprising of 3 domains, each domain has a val and test split.

This Hugging Face dataset repo contains the following contents:

datasets/facebook/SACo-VEval/tree/main/
├── annotation/
│   ├── saco_veval_sav_test.json
│   ├── saco_veval_sav_val.json
│   ├── saco_veval_smartglasses_test.json
│   ├── saco_veval_smartglasses_val.json
│   ├── saco_veval_yt1b_test.json
│   ├── saco_veval_yt1b_val.json
└── media/
    ├── saco_sg.tar.gz
    └── yt1b_start_end_time.json
  • annotation
    • all the GT json files
  • media
    • saco_sg.tar.gz: the preprocessed JPEGImages for SA-Co/VEval - SmartGlasses
    • yt1b_start_end_time.json: the Youtube video ids and the start and end time used in SA-Co/VEval - YT-Temporal-1B

More detail to prepare the complete SA-Co/VEval Dataset can be found in the SAM 3 Github.

Annotation Format

The format is similar to the YTVIS format.

In the annotation json, e.g. saco_veval_sav_test.json there are 5 fields:

  • info:
    • A dict containing the dataset info
    • E.g. {'version': 'v1', 'date': '2025-09-24', 'description': 'SA-Co/VEval SA-V Test'}
  • videos
    • A list of videos that are used in the current annotation json
    • It contains {id, video_name, file_names, height, width, length}
  • annotations
    • A list of positive masklets and their related info
    • It contains {id, segmentations, bboxes, areas, iscrowd, video_id, height, width, category_id, noun_phrase}
      • video_id should match to the videos - id field above
      • category_id should match to the categories - id field below
      • segmentations is a list of RLE
  • categories
    • A globally used noun phrase id map, which is true across all 3 domains.
    • It contains {id, name}
      • name is the noun phrase
  • video_np_pairs
    • A list of video-np pairs, including both positive and negative used in the current annotation json
    • It contains {id, video_id, category_id, noun_phrase, num_masklets}
      • video_id should match the videos - id above
      • category_id should match the categories - id above
      • when num_masklets > 0 it is a positive video-np pair, and the presenting masklets can be found in the annotations field
      • when num_masklets = 0 it is a negative video-np pair, meaning no masklet presenting at all
data {
    "info": info
    "videos": [video]
    "annotations": [annotation]
    "categories": [category]
    "video_np_pairs": [video_np_pair]
}
video {
    "id": int
    "video_name": str  # e.g. sav_000000
    "file_names": List[str]
    "height": int
    "width": width
    "length": length
}
annotation {
    "id": int
    "segmentations": List[RLE]
    "bboxes": List[List[int, int, int, int]]
    "areas": List[int]
    "iscrowd": int
    "video_id": str
    "height": int
    "width": int
    "category_id": int
    "noun_phrase": str
}
category {
    "id": int
    "name": str
}
video_np_pair {
    "id": int
    "video_id": str
    "category_id": int
    "noun_phrase": str
    "num_masklets" int
}

SAM 3 Github sam3/examples/saco_veval_vis_example.ipynb shows some examples of the data format and data visualization.
Downloads last month
-
Number of rows:
2
Total file size:
32.2 GB

Collection including ml-resources/SACo-VEval