10M SAM

The original json was obtained from v1.1.0, just with the RESOLUTION information added.

The format of image annotation file is as follows.

[
  {
    "path": "00168/001680102.jpg",
    "cap": [
      "xxxxx."
    ],
    "resolution": {
      "height": 512,
      "width": 683
    }
  },
  ...
]

6M HQ Panda70m

The format of video annotation file is as follows. Each element's path follows the structure: part_x/youtube_id/youtube_id_segment_i.mp4. Here, part_x is our custom organizational folder, which can be customed according to your download path. The youtube_id and segment_i can be obtained from the original annotation file.

[
  {
    "path": "panda70m_part_5565/qLqjjDhhD5Q/qLqjjDhhD5Q_segment_0.mp4",
    "cap": [
      "A man and a woman are sitting down on a news anchor talking to each other."
    ],
    "resolution": {
      "height": 720,
      "width": 1280
    },
    "fps": 29.97002997002997,
    "duration": 11.444767
  },
  ...
]

100k HQ data

The original data was obtained from v1.1.0. We reorganized captions.