{
  "id": 613115,
  "title": "Regarding the content of the dataset",
  "url": "/competitions/MABe-mouse-behavior-detection/discussion/613115",
  "author_name": "",
  "post_date": "2025-10-24T09:42:21.783694100Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hey, everyone! I just joined this competition and I have thoroughly read the competition introduction. Intuitively, I think this might be a competition related to the field of computer vision. However, after observing the dataset, I found that the video information (tracking) seems to be stored in the <strong>parquet</strong> format, which means do I only need to handle matrix data rather than the original video?</p>",
  "messages": [
    {
      "id": "3306312",
      "postDate": "10/24/2025 09:42:21",
      "content": "<p>Hey, everyone! I just joined this competition and I have thoroughly read the competition introduction. Intuitively, I think this might be a competition related to the field of computer vision. However, after observing the dataset, I found that the video information (tracking) seems to be stored in the <strong>parquet</strong> format, which means do I only need to handle matrix data rather than the original video?</p>",
      "rawMarkdown": "Hey, everyone! I just joined this competition and I have thoroughly read the competition introduction. Intuitively, I think this might be a competition related to the field of computer vision. However, after observing the dataset, I found that the video information (tracking) seems to be stored in the **parquet** format, which means do I only need to handle matrix data rather than the original video?",
      "votes": null
    },
    {
      "id": "3306464",
      "postDate": "10/24/2025 15:29:57",
      "content": "<p>Welcome! So computer vision is definitely involved in the study of animal behavior, but for this competition we have already done that step for you- we've taken the raw videos of interacting animals and performed markerless pose estimation (mostly using SLEAP or DeepLabCut), so that the data provided to teams is just the tracked poses of animals, from which they need to infer human-labeled behaviors. This means that the core of the challenge is more about time series modeling than computer vision. Hope that helps!</p>",
      "rawMarkdown": "Welcome! So computer vision is definitely involved in the study of animal behavior, but for this competition we have already done that step for you- we've taken the raw videos of interacting animals and performed markerless pose estimation (mostly using SLEAP or DeepLabCut), so that the data provided to teams is just the tracked poses of animals, from which they need to infer human-labeled behaviors. This means that the core of the challenge is more about time series modeling than computer vision. Hope that helps!",
      "votes": null
    },
    {
      "id": "3306465",
      "postDate": "10/24/2025 15:35:52",
      "content": "<p>Thank you very much for your reply, I think I understand this dataset.</p>",
      "rawMarkdown": "Thank you very much for your reply, I think I understand this dataset.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3306464,
      "author_name": "annkennedy05",
      "author_url": "",
      "post_date": "10/24/2025 15:29:57",
      "content": "<p>Welcome! So computer vision is definitely involved in the study of animal behavior, but for this competition we have already done that step for you- we've taken the raw videos of interacting animals and performed markerless pose estimation (mostly using SLEAP or DeepLabCut), so that the data provided to teams is just the tracked poses of animals, from which they need to infer human-labeled behaviors. This means that the core of the challenge is more about time series modeling than computer vision. Hope that helps!</p>",
      "votes": null,
      "replies": [
        {
          "id": 3306465,
          "author_name": "yichiisallyouneed",
          "author_url": "",
          "post_date": "10/24/2025 15:35:52",
          "content": "<p>Thank you very much for your reply, I think I understand this dataset.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3306312": "Hey, everyone! I just joined this competition and I have thoroughly read the competition introduction. Intuitively, I think this might be a competition related to the field of computer vision. However, after observing the dataset, I found that the video information (tracking) seems to be stored in the **parquet** format, which means do I only need to handle matrix data rather than the original video?",
    "3306464": "Welcome! So computer vision is definitely involved in the study of animal behavior, but for this competition we have already done that step for you- we've taken the raw videos of interacting animals and performed markerless pose estimation (mostly using SLEAP or DeepLabCut), so that the data provided to teams is just the tracked poses of animals, from which they need to infer human-labeled behaviors. This means that the core of the challenge is more about time series modeling than computer vision. Hope that helps!",
    "3306465": "Thank you very much for your reply, I think I understand this dataset."
  },
  "source": "meta"
}