{
  "id": 66386,
  "title": "Strategies to deal with grossly wrong predictions",
  "url": "/competitions/rsna-pneumonia-detection-challenge/discussion/66386",
  "author_name": "Eddie",
  "post_date": "2018-09-21T06:17:49.830000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Just looking to see what you guys are doing to deal with predictions that are very wrong.  Occasionally I get bounding boxes that are significantly larger than the target area and sometimes large ones in areas where there is no target.  I'm fairly new to ML and would love to hear some strategies to try and take these bad predictions and train the network to learn from them.</p>",
  "messages": [
    {
      "id": 391005,
      "postDate": "2018-09-21T06:17:49.830Z",
      "content": "<p>Just looking to see what you guys are doing to deal with predictions that are very wrong.  Occasionally I get bounding boxes that are significantly larger than the target area and sometimes large ones in areas where there is no target.  I'm fairly new to ML and would love to hear some strategies to try and take these bad predictions and train the network to learn from them.</p>",
      "rawMarkdown": "Just looking to see what you guys are doing to deal with predictions that are very wrong.  Occasionally I get bounding boxes that are significantly larger than the target area and sometimes large ones in areas where there is no target.  I'm fairly new to ML and would love to hear some strategies to try and take these bad predictions and train the network to learn from them."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "391005": "Just looking to see what you guys are doing to deal with predictions that are very wrong.  Occasionally I get bounding boxes that are significantly larger than the target area and sometimes large ones in areas where there is no target.  I'm fairly new to ML and would love to hear some strategies to try and take these bad predictions and train the network to learn from them."
  }
}