{
  "id": 588236,
  "title": "Two tags in a single video",
  "url": "/competitions/cupybara/discussion/588236",
  "author_name": "",
  "post_date": "2025-07-05T06:09:56.208810900Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi, I have a question regarding labeling videos where multiple species are clearly visible.</p>\n<p>In video 03260468.mp4, there is a large insect flying directly in front of the camera, very visible, while an armadillo is also present in the background, partially hidden behind some branches.</p>\n<p>Since both species are clearly identifiable, which one should we choose as the correct label?<br>\nShould we prioritize the species that is most visible and central, or the one that is ecologically more relevant, even if it appears in the background?</p>",
  "messages": [
    {
      "id": "3241732",
      "postDate": "07/05/2025 06:09:56",
      "content": "<p>Hi, I have a question regarding labeling videos where multiple species are clearly visible.</p>\n<p>In video 03260468.mp4, there is a large insect flying directly in front of the camera, very visible, while an armadillo is also present in the background, partially hidden behind some branches.</p>\n<p>Since both species are clearly identifiable, which one should we choose as the correct label?<br>\nShould we prioritize the species that is most visible and central, or the one that is ecologically more relevant, even if it appears in the background?</p>",
      "rawMarkdown": "Hi, I have a question regarding labeling videos where multiple species are clearly visible.\n\nIn video 03260468.mp4, there is a large insect flying directly in front of the camera, very visible, while an armadillo is also present in the background, partially hidden behind some branches.\n\nSince both species are clearly identifiable, which one should we choose as the correct label?\nShould we prioritize the species that is most visible and central, or the one that is ecologically more relevant, even if it appears in the background?",
      "votes": null
    },
    {
      "id": "3242561",
      "postDate": "07/06/2025 04:29:04",
      "content": "<p>Similar situations occur in videos 01100038.mp4 and 01120039.mp4, where more than one species from the class list is present and identifiable, but not equally prominent.</p>",
      "rawMarkdown": "Similar situations occur in videos 01100038.mp4 and 01120039.mp4, where more than one species from the class list is present and identifiable, but not equally prominent.",
      "votes": null
    },
    {
      "id": "3244928",
      "postDate": "07/08/2025 17:15:55",
      "content": "<p>Hi Axel,</p>\n<p>For any given video, the correct label should correspond to the most clearly identifiable and non-ambiguous animal present. If one animal is in focus, prominent, and clearly visible, that should be its label.</p>\n<p>A Note on Training Strategy:<br>\nIn a few rare cases, you might encounter a video where this rule is difficult to apply. How you handle these ambiguous examples in your own training process is a key part of the competition. The optimal strategy will be tied to your particular algorithm. For example, you might consider:</p>\n<p>Dropping an ambiguous video from your training set to prevent your model from learning confusing information.</p>\n<p>Duplicating the video in your training data and assigning a different valid label to each instance as a form of data augmentation.</p>\n<p>Ultimately, you have the flexibility to decide what will work best for your model. The goal is to build an algorithm that is robust and performs well on the hidden test set.</p>",
      "rawMarkdown": "Hi Axel,\n\nFor any given video, the correct label should correspond to the most clearly identifiable and non-ambiguous animal present. If one animal is in focus, prominent, and clearly visible, that should be its label.\n\nA Note on Training Strategy:\nIn a few rare cases, you might encounter a video where this rule is difficult to apply. How you handle these ambiguous examples in your own training process is a key part of the competition. The optimal strategy will be tied to your particular algorithm. For example, you might consider:\n\nDropping an ambiguous video from your training set to prevent your model from learning confusing information.\n\nDuplicating the video in your training data and assigning a different valid label to each instance as a form of data augmentation.\n\nUltimately, you have the flexibility to decide what will work best for your model. The goal is to build an algorithm that is robust and performs well on the hidden test set.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3242561,
      "author_name": "kitsukipho3nix",
      "author_url": "",
      "post_date": "07/06/2025 04:29:04",
      "content": "<p>Similar situations occur in videos 01100038.mp4 and 01120039.mp4, where more than one species from the class list is present and identifiable, but not equally prominent.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3244928,
      "author_name": "galvarez",
      "author_url": "",
      "post_date": "07/08/2025 17:15:55",
      "content": "<p>Hi Axel,</p>\n<p>For any given video, the correct label should correspond to the most clearly identifiable and non-ambiguous animal present. If one animal is in focus, prominent, and clearly visible, that should be its label.</p>\n<p>A Note on Training Strategy:<br>\nIn a few rare cases, you might encounter a video where this rule is difficult to apply. How you handle these ambiguous examples in your own training process is a key part of the competition. The optimal strategy will be tied to your particular algorithm. For example, you might consider:</p>\n<p>Dropping an ambiguous video from your training set to prevent your model from learning confusing information.</p>\n<p>Duplicating the video in your training data and assigning a different valid label to each instance as a form of data augmentation.</p>\n<p>Ultimately, you have the flexibility to decide what will work best for your model. The goal is to build an algorithm that is robust and performs well on the hidden test set.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3241732": "Hi, I have a question regarding labeling videos where multiple species are clearly visible.\n\nIn video 03260468.mp4, there is a large insect flying directly in front of the camera, very visible, while an armadillo is also present in the background, partially hidden behind some branches.\n\nSince both species are clearly identifiable, which one should we choose as the correct label?\nShould we prioritize the species that is most visible and central, or the one that is ecologically more relevant, even if it appears in the background?",
    "3242561": "Similar situations occur in videos 01100038.mp4 and 01120039.mp4, where more than one species from the class list is present and identifiable, but not equally prominent.",
    "3244928": "Hi Axel,\n\nFor any given video, the correct label should correspond to the most clearly identifiable and non-ambiguous animal present. If one animal is in focus, prominent, and clearly visible, that should be its label.\n\nA Note on Training Strategy:\nIn a few rare cases, you might encounter a video where this rule is difficult to apply. How you handle these ambiguous examples in your own training process is a key part of the competition. The optimal strategy will be tied to your particular algorithm. For example, you might consider:\n\nDropping an ambiguous video from your training set to prevent your model from learning confusing information.\n\nDuplicating the video in your training data and assigning a different valid label to each instance as a form of data augmentation.\n\nUltimately, you have the flexibility to decide what will work best for your model. The goal is to build an algorithm that is robust and performs well on the hidden test set."
  },
  "source": "meta"
}