{
  "id": 608753,
  "title": "A note on sleeping mice 🐭💤",
  "url": "/competitions/MABe-mouse-behavior-detection/discussion/608753",
  "author_name": "Ann Kennedy",
  "post_date": "2025-09-22T02:55:21.241000",
  "votes": 53,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi all, it has been great to see people starting to explore the dataset! Thank you to those who have asked questions, we will fix a few issues shortly.</p>\n<p>One point that may affect how you use the data: the two MABe22 datasets are collections of one-minute clips sampled at random from a group of experiments that recorded mouse behavior 24 hours a day for several days. The upshot of this is that mice sleep- so some of these clips are pretty uneventful! Furthermore, mice also love to sleep in a big pile called a huddle, which makes pose estimation during these times quite noisy.</p>\n<p>If you would like to exclude these clips from your analysis, simply check the \"mouse1 condition\" field of the csv: if it says <strong>\"lights on\"</strong> then the mice (who are nocturnal) are more likely to be <strong>sleeping</strong>. Or you can look at the tracking data itself and just toss out clips where there's no movement.</p>",
  "messages": [
    {
      "id": 3292517,
      "postDate": "2025-09-22T02:55:21.243Z",
      "content": "<p>Hi all, it has been great to see people starting to explore the dataset! Thank you to those who have asked questions, we will fix a few issues shortly.</p>\n<p>One point that may affect how you use the data: the two MABe22 datasets are collections of one-minute clips sampled at random from a group of experiments that recorded mouse behavior 24 hours a day for several days. The upshot of this is that mice sleep- so some of these clips are pretty uneventful! Furthermore, mice also love to sleep in a big pile called a huddle, which makes pose estimation during these times quite noisy.</p>\n<p>If you would like to exclude these clips from your analysis, simply check the \"mouse1 condition\" field of the csv: if it says <strong>\"lights on\"</strong> then the mice (who are nocturnal) are more likely to be <strong>sleeping</strong>. Or you can look at the tracking data itself and just toss out clips where there's no movement.</p>",
      "rawMarkdown": "Hi all, it has been great to see people starting to explore the dataset! Thank you to those who have asked questions, we will fix a few issues shortly.\n\nOne point that may affect how you use the data: the two MABe22 datasets are collections of one-minute clips sampled at random from a group of experiments that recorded mouse behavior 24 hours a day for several days. The upshot of this is that mice sleep- so some of these clips are pretty uneventful! Furthermore, mice also love to sleep in a big pile called a huddle, which makes pose estimation during these times quite noisy.\n\nIf you would like to exclude these clips from your analysis, simply check the \"mouse1 condition\" field of the csv: if it says **\"lights on\"** then the mice (who are nocturnal) are more likely to be **sleeping**. Or you can look at the tracking data itself and just toss out clips where there's no movement.",
      "votes": 52
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3292517": "Hi all, it has been great to see people starting to explore the dataset! Thank you to those who have asked questions, we will fix a few issues shortly.\n\nOne point that may affect how you use the data: the two MABe22 datasets are collections of one-minute clips sampled at random from a group of experiments that recorded mouse behavior 24 hours a day for several days. The upshot of this is that mice sleep- so some of these clips are pretty uneventful! Furthermore, mice also love to sleep in a big pile called a huddle, which makes pose estimation during these times quite noisy.\n\nIf you would like to exclude these clips from your analysis, simply check the \"mouse1 condition\" field of the csv: if it says **\"lights on\"** then the mice (who are nocturnal) are more likely to be **sleeping**. Or you can look at the tracking data itself and just toss out clips where there's no movement."
  }
}