{
  "id": 77165,
  "title": "Does anyone consider using a attention mechanism?",
  "url": "/competitions/histopathologic-cancer-detection/discussion/77165",
  "author_name": "",
  "post_date": "2019-01-10T06:50:54.055488700Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Does anyone consider using a attention mechanism?</p>",
  "messages": [
    {
      "id": "453412",
      "postDate": "01/10/2019 06:50:54",
      "content": "<p>Does anyone consider using a attention mechanism?</p>",
      "rawMarkdown": "Does anyone consider using a attention mechanism?",
      "votes": null
    },
    {
      "id": "456685",
      "postDate": "01/16/2019 09:52:26",
      "content": "<p>I tried to crop the images from the central part and train the network but the results were worse!</p>",
      "rawMarkdown": "I tried to crop the images from the central part and train the network but the results were worse!",
      "votes": null
    },
    {
      "id": "460309",
      "postDate": "01/23/2019 12:22:54",
      "content": "<p>me too......</p>",
      "rawMarkdown": "me too......",
      "votes": null
    },
    {
      "id": "460693",
      "postDate": "01/24/2019 08:53:43",
      "content": "<p>I tried STN, but without any profit</p>",
      "rawMarkdown": "I tried STN, but without any profit",
      "votes": null
    },
    {
      "id": "466097",
      "postDate": "02/04/2019 17:30:19",
      "content": "<p>What I can imagine is that some images are from the same macroscopic tissue/patient due to  'whole slide imaging' used (maybe even having another classifier when taking a look the image grid of the tissue sample). Hence cancer tissue in one image probably means cancer tissue in another one and so the samples are not independent. If we had the tissue map we could check...</p>",
      "rawMarkdown": "What I can imagine is that some images are from the same macroscopic tissue/patient due to  'whole slide imaging' used (maybe even having another classifier when taking a look the image grid of the tissue sample). Hence cancer tissue in one image probably means cancer tissue in another one and so the samples are not independent. If we had the tissue map we could check...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 456685,
      "author_name": "masih4",
      "author_url": "",
      "post_date": "01/16/2019 09:52:26",
      "content": "<p>I tried to crop the images from the central part and train the network but the results were worse!</p>",
      "votes": null,
      "replies": [
        {
          "id": 460309,
          "author_name": "chestnut111",
          "author_url": "",
          "post_date": "01/23/2019 12:22:54",
          "content": "<p>me too......</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 466097,
          "author_name": "wappyotoole",
          "author_url": "",
          "post_date": "02/04/2019 17:30:19",
          "content": "<p>What I can imagine is that some images are from the same macroscopic tissue/patient due to  'whole slide imaging' used (maybe even having another classifier when taking a look the image grid of the tissue sample). Hence cancer tissue in one image probably means cancer tissue in another one and so the samples are not independent. If we had the tissue map we could check...</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 460693,
      "author_name": "anikishaev",
      "author_url": "",
      "post_date": "01/24/2019 08:53:43",
      "content": "<p>I tried STN, but without any profit</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "453412": "Does anyone consider using a attention mechanism?",
    "456685": "I tried to crop the images from the central part and train the network but the results were worse!",
    "460309": "me too......",
    "460693": "I tried STN, but without any profit",
    "466097": "What I can imagine is that some images are from the same macroscopic tissue/patient due to  'whole slide imaging' used (maybe even having another classifier when taking a look the image grid of the tissue sample). Hence cancer tissue in one image probably means cancer tissue in another one and so the samples are not independent. If we had the tissue map we could check..."
  },
  "source": "meta"
}