{
  "id": 375751,
  "title": "Manually Refining/Creating Data",
  "url": "/competitions/nfl-player-contact-detection/discussion/375751",
  "author_name": "",
  "post_date": "2023-01-03T09:55:02.489311600Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p><strong>[1]</strong> There are oblivious mistakes in [test|train]_baseline_helmets.csv files. Are we allowed to refine these files by removing falsely detected helmets or adding undetected ones?</p>\n<p><strong>[2]</strong> Addition to the refining the existing data, are we allowed to manually create segmentation maps to train our DL models?</p>",
  "messages": [
    {
      "id": "2084226",
      "postDate": "01/03/2023 09:55:02",
      "content": "<p><strong>[1]</strong> There are oblivious mistakes in [test|train]_baseline_helmets.csv files. Are we allowed to refine these files by removing falsely detected helmets or adding undetected ones?</p>\n<p><strong>[2]</strong> Addition to the refining the existing data, are we allowed to manually create segmentation maps to train our DL models?</p>",
      "rawMarkdown": "**[1]** There are oblivious mistakes in [test|train]_baseline_helmets.csv files. Are we allowed to refine these files by removing falsely detected helmets or adding undetected ones?\n\n**[2]** Addition to the refining the existing data, are we allowed to manually create segmentation maps to train our DL models?",
      "votes": null
    },
    {
      "id": "2084488",
      "postDate": "01/03/2023 14:46:43",
      "content": "<p>Thanks for asking these questions <a href=\"https://www.kaggle.com/khanyorko\" target=\"_blank\">@khanyorko</a>.</p>\n<ol>\n<li><p>The baseline helmets is the output of a player assignment model and is not expected to be perfect. You are allowed to modify the training data in any way you find helpful. However, note that you will not have access to the test baseline helmet file. The test helmet file will only be seen by your model when your notebook is submitted.</p></li>\n<li><p>It is absolutely allowed to create any model to improve your predictions (segmentation maps or anything else).</p></li>\n</ol>",
      "rawMarkdown": "Thanks for asking these questions @khanyorko.\n\n1. The baseline helmets is the output of a player assignment model and is not expected to be perfect. You are allowed to modify the training data in any way you find helpful. However, note that you will not have access to the test baseline helmet file. The test helmet file will only be seen by your model when your notebook is submitted.\n\n2. It is absolutely allowed to create any model to improve your predictions (segmentation maps or anything else).",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2084488,
      "author_name": "robikscube",
      "author_url": "",
      "post_date": "01/03/2023 14:46:43",
      "content": "<p>Thanks for asking these questions <a href=\"https://www.kaggle.com/khanyorko\" target=\"_blank\">@khanyorko</a>.</p>\n<ol>\n<li><p>The baseline helmets is the output of a player assignment model and is not expected to be perfect. You are allowed to modify the training data in any way you find helpful. However, note that you will not have access to the test baseline helmet file. The test helmet file will only be seen by your model when your notebook is submitted.</p></li>\n<li><p>It is absolutely allowed to create any model to improve your predictions (segmentation maps or anything else).</p></li>\n</ol>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2084226": "**[1]** There are oblivious mistakes in [test|train]_baseline_helmets.csv files. Are we allowed to refine these files by removing falsely detected helmets or adding undetected ones?\n\n**[2]** Addition to the refining the existing data, are we allowed to manually create segmentation maps to train our DL models?",
    "2084488": "Thanks for asking these questions @khanyorko.\n\n1. The baseline helmets is the output of a player assignment model and is not expected to be perfect. You are allowed to modify the training data in any way you find helpful. However, note that you will not have access to the test baseline helmet file. The test helmet file will only be seen by your model when your notebook is submitted.\n\n2. It is absolutely allowed to create any model to improve your predictions (segmentation maps or anything else)."
  },
  "source": "meta"
}