{
  "id": 104689,
  "title": "Why can't we get good scores with only old competition dataset?",
  "url": "/competitions/aptos2019-blindness-detection/discussion/104689",
  "author_name": "",
  "post_date": "2019-08-18T13:18:09.745611700Z",
  "votes": 10,
  "comment_count": 1,
  "views": 0,
  "content": "<p>If the old competition data set and the current competition data set were created in the same way and annotated in the same way, we could get a good score just by using the old competition data set. \nI think there is any difference between the two datasets. And I think we should consider the difference when using old competition dataset. What do you think?</p>",
  "messages": [
    {
      "id": "602023",
      "postDate": "08/18/2019 13:18:09",
      "content": "<p>If the old competition data set and the current competition data set were created in the same way and annotated in the same way, we could get a good score just by using the old competition data set. \nI think there is any difference between the two datasets. And I think we should consider the difference when using old competition dataset. What do you think?</p>",
      "rawMarkdown": "If the old competition data set and the current competition data set were created in the same way and annotated in the same way, we could get a good score just by using the old competition data set. \nI think there is any difference between the two datasets. And I think we should consider the difference when using old competition dataset. What do you think?",
      "votes": null
    },
    {
      "id": "604987",
      "postDate": "08/21/2019 23:59:24",
      "content": "<p>It minds me this article. <a href=\"https://healthitanalytics.com/news/deep-learning-for-medical-imaging-fares-poorly-on-external-data\">https://healthitanalytics.com/news/deep-learning-for-medical-imaging-fares-poorly-on-external-data</a></p>",
      "rawMarkdown": "It minds me this article. https://healthitanalytics.com/news/deep-learning-for-medical-imaging-fares-poorly-on-external-data",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 604987,
      "author_name": "chikim",
      "author_url": "",
      "post_date": "08/21/2019 23:59:24",
      "content": "<p>It minds me this article. <a href=\"https://healthitanalytics.com/news/deep-learning-for-medical-imaging-fares-poorly-on-external-data\">https://healthitanalytics.com/news/deep-learning-for-medical-imaging-fares-poorly-on-external-data</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "602023": "If the old competition data set and the current competition data set were created in the same way and annotated in the same way, we could get a good score just by using the old competition data set. \nI think there is any difference between the two datasets. And I think we should consider the difference when using old competition dataset. What do you think?",
    "604987": "It minds me this article. https://healthitanalytics.com/news/deep-learning-for-medical-imaging-fares-poorly-on-external-data"
  },
  "source": "meta"
}