{
  "id": 110062,
  "title": "5, 6, 7, 31 or 32 classes?",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/110062",
  "author_name": "Daryl Brach",
  "post_date": "2019-09-24T13:13:02.143000",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>As we have seen in other posts, it is pretty straight forward that there are 5 or 6 classes depending on how you treat 'any'. </p>\n\n<p>| Class Type    Count\n1   epidural    2761\n2   intraparenchymal    32564\n3   intraventricular    23766\n4   subarachnoid    32122\n5   subdural    42496\n6   any 97103\n7      none \nHowever, when building a model, the issue of number of classes came up. How many classes or combinations of classes are there? How will it affect the overall classification of the images. So, after some Excel gymnastics, here are the class combinations and instance counts. </p>\n\n<p>Type    Count\nepidural    1497\nepidural_intraparenchymal   233\nepidural_intraparenchymal_intraventricular  45\nepidural_intraparenchymal_intraventricular_subarachnoid 36\nepidural_intraparenchymal_intraventricular_subarachnoid_subdural    20\nepidural_intraparenchymal_intraventricular_subdural 19\nepidural_intraparenchymal_subarachnoid  75\nepidural_intraparenchymal_subarachnoid_subdural 42\nepidural_intraparenchymal_subdural  86\nepidural_intraventricular   39\nepidural_intraventricular_subarachnoid  11\nepidural_intraventricular_subarachnoid_subdural 29\nepidural_intraventricular_subdural  17\nepidural_subarachnoid   180\nepidural_subarachnoid_subdural  91\nepidural_subdural   341\nintraparenchymal    13932\nintraparenchymal_intraventricular   6330\nintraparenchymal_intraventricular_subarachnoid  1761\nintraparenchymal_intraventricular_subarachnoid_subdural 759\nintraparenchymal_intraventricular_subdural  646\nintraparenchymal_subarachnoid   3611\nintraparenchymal_subarachnoid_subdural  2017\nintraparenchymal_subdural   2952\nintraventricular    8914\nintraventricular_subarachnoid   3226\nintraventricular_subarachnoid_subdural  893\nintraventricular_subdural   1021\nsubarachnoid    14717\nsubarachnoid_subdural   4654\nsubdural    28909\nnone\nTotal       97103</p>\n\n<p>What do you think?</p>",
  "messages": [
    {
      "id": 633407,
      "postDate": "2019-09-24T21:37:18.197Z",
      "content": "<p>no need to think just query <strong>multi-label classification</strong> and you will find your answer</p>",
      "rawMarkdown": "no need to think just query **multi-label classification** and you will find your answer",
      "votes": 1,
      "replies": [
        {
          "id": 636345,
          "postDate": "2019-09-29T10:33:10.190Z",
          "content": "<p>Thank you! I overthought this.</p>",
          "rawMarkdown": "Thank you! I overthought this."
        }
      ]
    },
    {
      "id": 633238,
      "postDate": "2019-09-24T15:35:28.260Z",
      "content": "<p>wat</p>",
      "rawMarkdown": "wat"
    },
    {
      "id": 633129,
      "postDate": "2019-09-24T13:13:02.143Z",
      "content": "<p>As we have seen in other posts, it is pretty straight forward that there are 5 or 6 classes depending on how you treat 'any'. </p>\n\n<p>| Class Type    Count\n1   epidural    2761\n2   intraparenchymal    32564\n3   intraventricular    23766\n4   subarachnoid    32122\n5   subdural    42496\n6   any 97103\n7      none \nHowever, when building a model, the issue of number of classes came up. How many classes or combinations of classes are there? How will it affect the overall classification of the images. So, after some Excel gymnastics, here are the class combinations and instance counts. </p>\n\n<p>Type    Count\nepidural    1497\nepidural_intraparenchymal   233\nepidural_intraparenchymal_intraventricular  45\nepidural_intraparenchymal_intraventricular_subarachnoid 36\nepidural_intraparenchymal_intraventricular_subarachnoid_subdural    20\nepidural_intraparenchymal_intraventricular_subdural 19\nepidural_intraparenchymal_subarachnoid  75\nepidural_intraparenchymal_subarachnoid_subdural 42\nepidural_intraparenchymal_subdural  86\nepidural_intraventricular   39\nepidural_intraventricular_subarachnoid  11\nepidural_intraventricular_subarachnoid_subdural 29\nepidural_intraventricular_subdural  17\nepidural_subarachnoid   180\nepidural_subarachnoid_subdural  91\nepidural_subdural   341\nintraparenchymal    13932\nintraparenchymal_intraventricular   6330\nintraparenchymal_intraventricular_subarachnoid  1761\nintraparenchymal_intraventricular_subarachnoid_subdural 759\nintraparenchymal_intraventricular_subdural  646\nintraparenchymal_subarachnoid   3611\nintraparenchymal_subarachnoid_subdural  2017\nintraparenchymal_subdural   2952\nintraventricular    8914\nintraventricular_subarachnoid   3226\nintraventricular_subarachnoid_subdural  893\nintraventricular_subdural   1021\nsubarachnoid    14717\nsubarachnoid_subdural   4654\nsubdural    28909\nnone\nTotal       97103</p>\n\n<p>What do you think?</p>",
      "rawMarkdown": "As we have seen in other posts, it is pretty straight forward that there are 5 or 6 classes depending on how you treat 'any'. \n\n| Class\tType\tCount\n1\tepidural\t2761\n2\tintraparenchymal\t32564\n3\tintraventricular\t23766\n4\tsubarachnoid\t32122\n5\tsubdural \t42496\n6\tany\t97103\n7      none \nHowever, when building a model, the issue of number of classes came up. How many classes or combinations of classes are there? How will it affect the overall classification of the images. So, after some Excel gymnastics, here are the class combinations and instance counts. \n\nType\tCount\nepidural\t1497\nepidural_intraparenchymal\t233\nepidural_intraparenchymal_intraventricular\t45\nepidural_intraparenchymal_intraventricular_subarachnoid\t36\nepidural_intraparenchymal_intraventricular_subarachnoid_subdural\t20\nepidural_intraparenchymal_intraventricular_subdural\t19\nepidural_intraparenchymal_subarachnoid\t75\nepidural_intraparenchymal_subarachnoid_subdural\t42\nepidural_intraparenchymal_subdural\t86\nepidural_intraventricular\t39\nepidural_intraventricular_subarachnoid\t11\nepidural_intraventricular_subarachnoid_subdural\t29\nepidural_intraventricular_subdural\t17\nepidural_subarachnoid\t180\nepidural_subarachnoid_subdural\t91\nepidural_subdural\t341\nintraparenchymal\t13932\nintraparenchymal_intraventricular\t6330\nintraparenchymal_intraventricular_subarachnoid\t1761\nintraparenchymal_intraventricular_subarachnoid_subdural\t759\nintraparenchymal_intraventricular_subdural\t646\nintraparenchymal_subarachnoid\t3611\nintraparenchymal_subarachnoid_subdural\t2017\nintraparenchymal_subdural\t2952\nintraventricular\t8914\nintraventricular_subarachnoid\t3226\nintraventricular_subarachnoid_subdural\t893\nintraventricular_subdural\t1021\nsubarachnoid\t14717\nsubarachnoid_subdural\t4654\nsubdural\t28909\nnone\nTotal\t\t97103\n\nWhat do you think?",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 633407,
      "author_name": "Miroslav Valan",
      "author_url": "",
      "post_date": "2019-09-24T21:37:18.197000",
      "content": "<p>no need to think just query <strong>multi-label classification</strong> and you will find your answer</p>",
      "votes": 1,
      "replies": [
        {
          "id": 636345,
          "author_name": "Daryl Brach",
          "author_url": "",
          "post_date": "2019-09-29T10:33:10.190000",
          "content": "<p>Thank you! I overthought this.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 633238,
      "author_name": "JM",
      "author_url": "",
      "post_date": "2019-09-24T15:35:28.260000",
      "content": "<p>wat</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "633407": "no need to think just query **multi-label classification** and you will find your answer",
    "633238": "wat",
    "633129": "As we have seen in other posts, it is pretty straight forward that there are 5 or 6 classes depending on how you treat 'any'. \n\n| Class\tType\tCount\n1\tepidural\t2761\n2\tintraparenchymal\t32564\n3\tintraventricular\t23766\n4\tsubarachnoid\t32122\n5\tsubdural \t42496\n6\tany\t97103\n7      none \nHowever, when building a model, the issue of number of classes came up. How many classes or combinations of classes are there? How will it affect the overall classification of the images. So, after some Excel gymnastics, here are the class combinations and instance counts. \n\nType\tCount\nepidural\t1497\nepidural_intraparenchymal\t233\nepidural_intraparenchymal_intraventricular\t45\nepidural_intraparenchymal_intraventricular_subarachnoid\t36\nepidural_intraparenchymal_intraventricular_subarachnoid_subdural\t20\nepidural_intraparenchymal_intraventricular_subdural\t19\nepidural_intraparenchymal_subarachnoid\t75\nepidural_intraparenchymal_subarachnoid_subdural\t42\nepidural_intraparenchymal_subdural\t86\nepidural_intraventricular\t39\nepidural_intraventricular_subarachnoid\t11\nepidural_intraventricular_subarachnoid_subdural\t29\nepidural_intraventricular_subdural\t17\nepidural_subarachnoid\t180\nepidural_subarachnoid_subdural\t91\nepidural_subdural\t341\nintraparenchymal\t13932\nintraparenchymal_intraventricular\t6330\nintraparenchymal_intraventricular_subarachnoid\t1761\nintraparenchymal_intraventricular_subarachnoid_subdural\t759\nintraparenchymal_intraventricular_subdural\t646\nintraparenchymal_subarachnoid\t3611\nintraparenchymal_subarachnoid_subdural\t2017\nintraparenchymal_subdural\t2952\nintraventricular\t8914\nintraventricular_subarachnoid\t3226\nintraventricular_subarachnoid_subdural\t893\nintraventricular_subdural\t1021\nsubarachnoid\t14717\nsubarachnoid_subdural\t4654\nsubdural\t28909\nnone\nTotal\t\t97103\n\nWhat do you think?"
  }
}