{
  "id": 104184,
  "title": "Aptos 2019 Training Images (Preprocessed and cropped using Ben's preprocessing)",
  "url": "/competitions/aptos2019-blindness-detection/discussion/104184",
  "author_name": "Quan",
  "post_date": "2019-08-15T02:30:42.616000",
  "votes": 0,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Shout out to Ben (Last competition's winner) for amazing preprocessing. His method effectively improves lighting condition across images. \nIn this dataset, I used his preprocessing method, crop excess graybars and resize to 224x224 images.\n<a href=\"https://www.kaggle.com/quandapro/training-images-224\">https://www.kaggle.com/quandapro/training-images-224</a></p>",
  "messages": [
    {
      "id": 599472,
      "postDate": "2019-08-15T02:40:23.333Z",
      "content": "<p>Thanks  man , can  you  share  your  processed  code  in   the   dataset  description    or  some kernel  ?   Two  weeks  ago , I  did  the  same  thing  like  you  ,  but  I  can  only  get  0.66  LB  score  ,   this   issue  makes  me  could only   train  in  2019's  competition  data  , I want  to  know  which  step  I  made  a  mistake (resize  or  saving) ,   so  can  I  have  a look    about  your  processing  code   if   it's  convenient   to  you  .\nThanks  for  your  post  again.</p>",
      "rawMarkdown": "Thanks  man , can  you  share  your  processed  code  in   the   dataset  description    or  some kernel  ?   Two  weeks  ago , I  did  the  same  thing  like  you  ,  but  I  can  only  get  0.66  LB  score  ,   this   issue  makes  me  could only   train  in  2019's  competition  data  , I want  to  know  which  step  I  made  a  mistake (resize  or  saving) ,   so  can  I  have  a look    about  your  processing  code   if   it's  convenient   to  you  .\nThanks  for  your  post  again.\n"
    },
    {
      "id": 599470,
      "postDate": "2019-08-15T02:30:42.617Z",
      "content": "<p>Shout out to Ben (Last competition's winner) for amazing preprocessing. His method effectively improves lighting condition across images. \nIn this dataset, I used his preprocessing method, crop excess graybars and resize to 224x224 images.\n<a href=\"https://www.kaggle.com/quandapro/training-images-224\">https://www.kaggle.com/quandapro/training-images-224</a></p>",
      "rawMarkdown": "Shout out to Ben (Last competition's winner) for amazing preprocessing. His method effectively improves lighting condition across images. \nIn this dataset, I used his preprocessing method, crop excess graybars and resize to 224x224 images.\nhttps://www.kaggle.com/quandapro/training-images-224"
    },
    {
      "id": 599477,
      "postDate": "2019-08-15T02:53:09.463Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 599472,
      "author_name": "哈尔的移动城堡",
      "author_url": "",
      "post_date": "2019-08-15T02:40:23.333000",
      "content": "<p>Thanks  man , can  you  share  your  processed  code  in   the   dataset  description    or  some kernel  ?   Two  weeks  ago , I  did  the  same  thing  like  you  ,  but  I  can  only  get  0.66  LB  score  ,   this   issue  makes  me  could only   train  in  2019's  competition  data  , I want  to  know  which  step  I  made  a  mistake (resize  or  saving) ,   so  can  I  have  a look    about  your  processing  code   if   it's  convenient   to  you  .\nThanks  for  your  post  again.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 599477,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-08-15T02:53:09.463000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "599472": "Thanks  man , can  you  share  your  processed  code  in   the   dataset  description    or  some kernel  ?   Two  weeks  ago , I  did  the  same  thing  like  you  ,  but  I  can  only  get  0.66  LB  score  ,   this   issue  makes  me  could only   train  in  2019's  competition  data  , I want  to  know  which  step  I  made  a  mistake (resize  or  saving) ,   so  can  I  have  a look    about  your  processing  code   if   it's  convenient   to  you  .\nThanks  for  your  post  again.\n",
    "599470": "Shout out to Ben (Last competition's winner) for amazing preprocessing. His method effectively improves lighting condition across images. \nIn this dataset, I used his preprocessing method, crop excess graybars and resize to 224x224 images.\nhttps://www.kaggle.com/quandapro/training-images-224",
    "599477": ""
  }
}