{
  "id": 680129,
  "title": "Data Set labelling",
  "url": "/competitions/accident/discussion/680129",
  "author_name": "",
  "post_date": "2026-03-05T22:49:05.999741300Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hey, I have a question about how the data sets work. I saw from an earlier discussion that you are allowed to use outside data sets as long as you are not labelling them yourself. Does this mean that we are not allowed to label the provided non-synthetic data (real-life accident example), and if it is an outside data set, is it fine if it comes labeled, just not self labeld</p>",
  "messages": [
    {
      "id": "3417655",
      "postDate": "03/05/2026 22:49:06",
      "content": "<p>Hey, I have a question about how the data sets work. I saw from an earlier discussion that you are allowed to use outside data sets as long as you are not labelling them yourself. Does this mean that we are not allowed to label the provided non-synthetic data (real-life accident example), and if it is an outside data set, is it fine if it comes labeled, just not self labeld</p>",
      "rawMarkdown": "Hey, I have a question about how the data sets work. I saw from an earlier discussion that you are allowed to use outside data sets as long as you are not labelling them yourself. Does this mean that we are not allowed to label the provided non-synthetic data (real-life accident example), and if it is an outside data set, is it fine if it comes labeled, just not self labeld",
      "votes": null
    },
    {
      "id": "3417674",
      "postDate": "03/06/2026 00:04:01",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/zookeeper9655\" target=\"_blank\">@zookeeper9655</a> ,</p>\n<p>Not sure if I completely understand the question, but I can confirm that:</p>\n<ol>\n<li>You cannot manually label the test dataset, i.e., non-synthetic data.</li>\n<li>You can use external data annotations. But given that some of the videos are included in other datasets, you can only use the annotations for training, not for heuristics or anything that would identify/recognize the test videos (or otherwise leak test information).</li>\n</ol>\n<p>Does it answer your question? :)</p>\n<p>Best,\nLukas</p>",
      "rawMarkdown": "Hi @zookeeper9655 ,\n\nNot sure if I completely understand the question, but I can confirm that:\n1. You cannot manually label the test dataset, i.e., non-synthetic data.\n2. You can use external data annotations. But given that some of the videos are included in other datasets, you can only use the annotations for training, not for heuristics or anything that would identify/recognize the test videos (or otherwise leak test information).\n\nDoes it answer your question? :)\n\nBest,\nLukas",
      "votes": null
    },
    {
      "id": "3417717",
      "postDate": "03/06/2026 01:51:51",
      "content": "<p>Yes, thank you so much.</p>\n<ol>\n<li>One can use labeled outside data, just avoid leaking data that would help one's model pass the test</li>\n<li>One cannot manually label any of the given test data</li>\n</ol>",
      "rawMarkdown": "Yes, thank you so much.\n1. One can use labeled outside data, just avoid leaking data that would help one's model pass the test\n2. One cannot manually label any of the given test data",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3417674,
      "author_name": "picekl",
      "author_url": "",
      "post_date": "03/06/2026 00:04:01",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/zookeeper9655\" target=\"_blank\">@zookeeper9655</a> ,</p>\n<p>Not sure if I completely understand the question, but I can confirm that:</p>\n<ol>\n<li>You cannot manually label the test dataset, i.e., non-synthetic data.</li>\n<li>You can use external data annotations. But given that some of the videos are included in other datasets, you can only use the annotations for training, not for heuristics or anything that would identify/recognize the test videos (or otherwise leak test information).</li>\n</ol>\n<p>Does it answer your question? :)</p>\n<p>Best,\nLukas</p>",
      "votes": null,
      "replies": [
        {
          "id": 3417717,
          "author_name": "zookeeper9655",
          "author_url": "",
          "post_date": "03/06/2026 01:51:51",
          "content": "<p>Yes, thank you so much.</p>\n<ol>\n<li>One can use labeled outside data, just avoid leaking data that would help one's model pass the test</li>\n<li>One cannot manually label any of the given test data</li>\n</ol>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3417655": "Hey, I have a question about how the data sets work. I saw from an earlier discussion that you are allowed to use outside data sets as long as you are not labelling them yourself. Does this mean that we are not allowed to label the provided non-synthetic data (real-life accident example), and if it is an outside data set, is it fine if it comes labeled, just not self labeld",
    "3417674": "Hi @zookeeper9655 ,\n\nNot sure if I completely understand the question, but I can confirm that:\n1. You cannot manually label the test dataset, i.e., non-synthetic data.\n2. You can use external data annotations. But given that some of the videos are included in other datasets, you can only use the annotations for training, not for heuristics or anything that would identify/recognize the test videos (or otherwise leak test information).\n\nDoes it answer your question? :)\n\nBest,\nLukas",
    "3417717": "Yes, thank you so much.\n1. One can use labeled outside data, just avoid leaking data that would help one's model pass the test\n2. One cannot manually label any of the given test data"
  },
  "source": "meta"
}