{
  "id": 248832,
  "title": "About Re-Labeling for train data",
  "url": "/competitions/siim-covid19-detection/discussion/248832",
  "author_name": "",
  "post_date": "2021-06-25T06:48:34.774106800Z",
  "votes": 8,
  "comment_count": 3,
  "views": 0,
  "content": "<p><a href=\"https://www.kaggle.com/juliaelliott\" target=\"_blank\">@juliaelliott</a> <br>\nI'm now wondering if in this competition manually re-labeling training dataset (for bbox or classes) is allowed or not.<br>\nfor some of the other comps, when train data are noisy, winners solutions sometimes contained re-labeling train data manually and trained with those data. (like sound classification competition).<br>\nIs it allowed for this competition too?</p>",
  "messages": [
    {
      "id": "1364708",
      "postDate": "06/25/2021 06:48:34",
      "content": "<p><a href=\"https://www.kaggle.com/juliaelliott\" target=\"_blank\">@juliaelliott</a> <br>\nI'm now wondering if in this competition manually re-labeling training dataset (for bbox or classes) is allowed or not.<br>\nfor some of the other comps, when train data are noisy, winners solutions sometimes contained re-labeling train data manually and trained with those data. (like sound classification competition).<br>\nIs it allowed for this competition too?</p>",
      "rawMarkdown": "juliaelliott \nI'm now wondering if in this competition manually re-labeling training dataset (for bbox or classes) is allowed or not.\nfor some of the other comps, when train data are noisy, winners solutions sometimes contained re-labeling train data manually and trained with those data. (like sound classification competition).\nIs it allowed for this competition too?",
      "votes": null
    },
    {
      "id": "1368868",
      "postDate": "06/29/2021 01:26:13",
      "content": "<p>This comp is difficult for re-labeling I think.</p>",
      "rawMarkdown": "This comp is difficult for re-labeling I think.",
      "votes": null
    },
    {
      "id": "1369902",
      "postDate": "06/29/2021 17:31:41",
      "content": "<p>Training data is fine to re-label. However, per the rules, the test set cannot be hand-labeled in any way.</p>",
      "rawMarkdown": "Training data is fine to re-label. However, per the rules, the test set cannot be hand-labeled in any way.",
      "votes": null
    },
    {
      "id": "1370070",
      "postDate": "06/29/2021 21:51:04",
      "content": "<p>Thank you Julia,<br>\nOk, We definitely do it for only train df, not for test if we re-label.</p>",
      "rawMarkdown": "Thank you Julia,\nOk, We definitely do it for only train df, not for test if we re-label.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1368868,
      "author_name": "mingtai",
      "author_url": "",
      "post_date": "06/29/2021 01:26:13",
      "content": "<p>This comp is difficult for re-labeling I think.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1369902,
      "author_name": "juliaelliott",
      "author_url": "",
      "post_date": "06/29/2021 17:31:41",
      "content": "<p>Training data is fine to re-label. However, per the rules, the test set cannot be hand-labeled in any way.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1370070,
          "author_name": "ryunosukeishizaki",
          "author_url": "",
          "post_date": "06/29/2021 21:51:04",
          "content": "<p>Thank you Julia,<br>\nOk, We definitely do it for only train df, not for test if we re-label.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1364708": "juliaelliott \nI'm now wondering if in this competition manually re-labeling training dataset (for bbox or classes) is allowed or not.\nfor some of the other comps, when train data are noisy, winners solutions sometimes contained re-labeling train data manually and trained with those data. (like sound classification competition).\nIs it allowed for this competition too?",
    "1368868": "This comp is difficult for re-labeling I think.",
    "1369902": "Training data is fine to re-label. However, per the rules, the test set cannot be hand-labeled in any way.",
    "1370070": "Thank you Julia,\nOk, We definitely do it for only train df, not for test if we re-label."
  },
  "source": "meta"
}