{
  "id": 95198,
  "title": "Discussion about the solutions",
  "url": "/competitions/iwildcam-2019-fgvc6/discussion/95198",
  "author_name": "eclipse",
  "post_date": "2019-06-10T17:12:47.090000",
  "votes": 6,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I was very, very lucky to get the 7th place in private LB.  The source code of my solution is available at <a href=\"https://github.com/Walleclipse/iWildCam_2019_FGVC6\">https://github.com/Walleclipse/iWildCam_2019_FGVC6</a> \nI found, my score is much worse than those of the top teams. I am eager to discuss with you the different solutions to this competition.  Does anyone consider the data shift (or domain shift) between train dataset and test dataset? Does anyone try to GAN or transfer learning with domain shift?  What other issues need to be considered?</p>",
  "messages": [
    {
      "id": 549463,
      "postDate": "2019-06-10T17:12:47.090Z",
      "content": "<p>I was very, very lucky to get the 7th place in private LB.  The source code of my solution is available at <a href=\"https://github.com/Walleclipse/iWildCam_2019_FGVC6\">https://github.com/Walleclipse/iWildCam_2019_FGVC6</a> \nI found, my score is much worse than those of the top teams. I am eager to discuss with you the different solutions to this competition.  Does anyone consider the data shift (or domain shift) between train dataset and test dataset? Does anyone try to GAN or transfer learning with domain shift?  What other issues need to be considered?</p>",
      "rawMarkdown": "I was very, very lucky to get the 7th place in private LB.  The source code of my solution is available at [https://github.com/Walleclipse/iWildCam\\_2019\\_FGVC6](https://github.com/Walleclipse/iWildCam_2019_FGVC6) \nI found, my score is much worse than those of the top teams. I am eager to discuss with you the different solutions to this competition.  Does anyone consider the data shift (or domain shift) between train dataset and test dataset? Does anyone try to GAN or transfer learning with domain shift?  What other issues need to be considered?",
      "votes": 6
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "549463": "I was very, very lucky to get the 7th place in private LB.  The source code of my solution is available at [https://github.com/Walleclipse/iWildCam\\_2019\\_FGVC6](https://github.com/Walleclipse/iWildCam_2019_FGVC6) \nI found, my score is much worse than those of the top teams. I am eager to discuss with you the different solutions to this competition.  Does anyone consider the data shift (or domain shift) between train dataset and test dataset? Does anyone try to GAN or transfer learning with domain shift?  What other issues need to be considered?"
  }
}