{
  "id": 576568,
  "title": "Video length different from 15 seconds",
  "url": "/competitions/cupybara/discussion/576568",
  "author_name": "",
  "post_date": "2025-05-05T20:55:35.668696400Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello, I stumbled upon an issue while exploring the dataset. Some videos inside the train set like 02070026.mp4 and 02090042.mp4 are 13s and 5s, respectively. I believe many more are shorter than 15s. Should we find a way to bypass this issue by cutting videos shorter or will the organizers provide a correction for the data? Thanks very much for the attention.</p>",
  "messages": [
    {
      "id": "3194412",
      "postDate": "05/05/2025 20:55:35",
      "content": "<p>Hello, I stumbled upon an issue while exploring the dataset. Some videos inside the train set like 02070026.mp4 and 02090042.mp4 are 13s and 5s, respectively. I believe many more are shorter than 15s. Should we find a way to bypass this issue by cutting videos shorter or will the organizers provide a correction for the data? Thanks very much for the attention.</p>",
      "rawMarkdown": "Hello, I stumbled upon an issue while exploring the dataset. Some videos inside the train set like 02070026.mp4 and 02090042.mp4 are 13s and 5s, respectively. I believe many more are shorter than 15s. Should we find a way to bypass this issue by cutting videos shorter or will the organizers provide a correction for the data? Thanks very much for the attention.",
      "votes": null
    },
    {
      "id": "3194989",
      "postDate": "05/06/2025 14:03:25",
      "content": "<p>Hi Tomas,</p>\n<p>Thanks for flagging this!</p>\n<p>You're right — while the majority of the videos in the dataset are 15 seconds long, some videos are shorter due to capture or upload issues from the trap cameras. These include examples like the ones you mentioned (13s, 5s, etc.).</p>\n<p>We won’t be modifying the dataset on our side, but you're welcome to handle this in your pipeline however you see fit. Two commonly used options are:</p>\n<ul>\n<li><p>Padding: Extending shorter videos to a fixed length (e.g., duplicating the last frame, zero-padding, etc.)</p></li>\n<li><p>Omitting: Skipping those samples if you prefer consistency in input length</p></li>\n</ul>\n<p>We encourage teams to document their approach in their final solution description if this ends up being a relevant factor in your pipeline.</p>\n<p>Remember that omitting or shortening the length of the videos will end up in loosing data.</p>\n<p>Let us know if you encounter anything else!</p>",
      "rawMarkdown": "Hi Tomas,\n\nThanks for flagging this!\n\nYou're right — while the majority of the videos in the dataset are 15 seconds long, some videos are shorter due to capture or upload issues from the trap cameras. These include examples like the ones you mentioned (13s, 5s, etc.).\n\nWe won’t be modifying the dataset on our side, but you're welcome to handle this in your pipeline however you see fit. Two commonly used options are:\n\n- Padding: Extending shorter videos to a fixed length (e.g., duplicating the last frame, zero-padding, etc.)\n\n- Omitting: Skipping those samples if you prefer consistency in input length\n\nWe encourage teams to document their approach in their final solution description if this ends up being a relevant factor in your pipeline.\n\nRemember that omitting or shortening the length of the videos will end up in loosing data.\n\nLet us know if you encounter anything else!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3194989,
      "author_name": "galvarez",
      "author_url": "",
      "post_date": "05/06/2025 14:03:25",
      "content": "<p>Hi Tomas,</p>\n<p>Thanks for flagging this!</p>\n<p>You're right — while the majority of the videos in the dataset are 15 seconds long, some videos are shorter due to capture or upload issues from the trap cameras. These include examples like the ones you mentioned (13s, 5s, etc.).</p>\n<p>We won’t be modifying the dataset on our side, but you're welcome to handle this in your pipeline however you see fit. Two commonly used options are:</p>\n<ul>\n<li><p>Padding: Extending shorter videos to a fixed length (e.g., duplicating the last frame, zero-padding, etc.)</p></li>\n<li><p>Omitting: Skipping those samples if you prefer consistency in input length</p></li>\n</ul>\n<p>We encourage teams to document their approach in their final solution description if this ends up being a relevant factor in your pipeline.</p>\n<p>Remember that omitting or shortening the length of the videos will end up in loosing data.</p>\n<p>Let us know if you encounter anything else!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3194412": "Hello, I stumbled upon an issue while exploring the dataset. Some videos inside the train set like 02070026.mp4 and 02090042.mp4 are 13s and 5s, respectively. I believe many more are shorter than 15s. Should we find a way to bypass this issue by cutting videos shorter or will the organizers provide a correction for the data? Thanks very much for the attention.",
    "3194989": "Hi Tomas,\n\nThanks for flagging this!\n\nYou're right — while the majority of the videos in the dataset are 15 seconds long, some videos are shorter due to capture or upload issues from the trap cameras. These include examples like the ones you mentioned (13s, 5s, etc.).\n\nWe won’t be modifying the dataset on our side, but you're welcome to handle this in your pipeline however you see fit. Two commonly used options are:\n\n- Padding: Extending shorter videos to a fixed length (e.g., duplicating the last frame, zero-padding, etc.)\n\n- Omitting: Skipping those samples if you prefer consistency in input length\n\nWe encourage teams to document their approach in their final solution description if this ends up being a relevant factor in your pipeline.\n\nRemember that omitting or shortening the length of the videos will end up in loosing data.\n\nLet us know if you encounter anything else!"
  },
  "source": "meta"
}