{
  "id": 110625,
  "title": "Some slices are missing labels",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/110625",
  "author_name": "Tom H.",
  "post_date": "2019-09-29T20:39:12.536000",
  "votes": 0,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Approximately 78,545 images (2,214 studies) are missing labels in stage_1_train.csv. Does anyone have information on this?</p>\n\n<p>Omitting all studies with slices that have missing labels is an option, but that's quite a few studies to throw out.</p>\n\n<p>Thanks</p>",
  "messages": [
    {
      "id": 636591,
      "postDate": "2019-09-29T20:58:34.117Z",
      "content": "<p>Those are the stage 1 test images. </p>",
      "rawMarkdown": "Those are the stage 1 test images. ",
      "votes": 2,
      "replies": [
        {
          "id": 636933,
          "postDate": "2019-09-30T11:42:14.380Z",
          "content": "<p>Thanks Ian! We aren't given labels on the test images? How are we supposed to evaluate accuracy? I thought the two folders (test/train) were split arbitrarily - for convenience's sake.</p>",
          "rawMarkdown": "Thanks Ian! We aren't given labels on the test images? How are we supposed to evaluate accuracy? I thought the two folders (test/train) were split arbitrarily - for convenience's sake."
        },
        {
          "id": 636966,
          "postDate": "2019-09-30T12:49:55.877Z",
          "content": "<p>You make predictions on the test images and submit them to the leaderboard. You can make your own local validation set by using a portion (10-20%) of the training data. </p>",
          "rawMarkdown": "You make predictions on the test images and submit them to the leaderboard. You can make your own local validation set by using a portion (10-20%) of the training data. ",
          "votes": 1
        }
      ]
    },
    {
      "id": 636584,
      "postDate": "2019-09-29T20:39:12.537Z",
      "content": "<p>Approximately 78,545 images (2,214 studies) are missing labels in stage_1_train.csv. Does anyone have information on this?</p>\n\n<p>Omitting all studies with slices that have missing labels is an option, but that's quite a few studies to throw out.</p>\n\n<p>Thanks</p>",
      "rawMarkdown": "Approximately 78,545 images (2,214 studies) are missing labels in stage_1_train.csv. Does anyone have information on this?\n\nOmitting all studies with slices that have missing labels is an option, but that's quite a few studies to throw out.\n\nThanks"
    }
  ],
  "comments": [
    {
      "id": 636591,
      "author_name": "Ian Pan",
      "author_url": "",
      "post_date": "2019-09-29T20:58:34.117000",
      "content": "<p>Those are the stage 1 test images. </p>",
      "votes": 2,
      "replies": [
        {
          "id": 636933,
          "author_name": "Tom H.",
          "author_url": "",
          "post_date": "2019-09-30T11:42:14.380000",
          "content": "<p>Thanks Ian! We aren't given labels on the test images? How are we supposed to evaluate accuracy? I thought the two folders (test/train) were split arbitrarily - for convenience's sake.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 636966,
          "author_name": "Ian Pan",
          "author_url": "",
          "post_date": "2019-09-30T12:49:55.877000",
          "content": "<p>You make predictions on the test images and submit them to the leaderboard. You can make your own local validation set by using a portion (10-20%) of the training data. </p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "636591": "Those are the stage 1 test images. ",
    "636584": "Approximately 78,545 images (2,214 studies) are missing labels in stage_1_train.csv. Does anyone have information on this?\n\nOmitting all studies with slices that have missing labels is an option, but that's quite a few studies to throw out.\n\nThanks"
  }
}