{
  "id": 254057,
  "title": "Negative Samples for EfficientDet?",
  "url": "/competitions/siim-covid19-detection/discussion/254057",
  "author_name": "Jaechan Lee",
  "post_date": "2021-07-20T00:13:52.924000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi, I am a new kaggler, and this is my first competition.</p>\n<p>I'm trying to implement EfficientDet, but I realized that EffficientDet per se has no way to handle images without bounding boxes (or \"none\" images), unlike YoloV5 and FRCNN which have special ways to process them. I might just use the 0 0 1 1 label but I feel like that might make my model perform worse.</p>\n<p>What should I do?</p>",
  "messages": [
    {
      "id": 1393826,
      "postDate": "2021-07-20T00:13:52.923Z",
      "content": "<p>Hi, I am a new kaggler, and this is my first competition.</p>\n<p>I'm trying to implement EfficientDet, but I realized that EffficientDet per se has no way to handle images without bounding boxes (or \"none\" images), unlike YoloV5 and FRCNN which have special ways to process them. I might just use the 0 0 1 1 label but I feel like that might make my model perform worse.</p>\n<p>What should I do?</p>",
      "rawMarkdown": "Hi, I am a new kaggler, and this is my first competition.\n\nI'm trying to implement EfficientDet, but I realized that EffficientDet per se has no way to handle images without bounding boxes (or \"none\" images), unlike YoloV5 and FRCNN which have special ways to process them. I might just use the 0 0 1 1 label but I feel like that might make my model perform worse.\n\nWhat should I do?",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1393826": "Hi, I am a new kaggler, and this is my first competition.\n\nI'm trying to implement EfficientDet, but I realized that EffficientDet per se has no way to handle images without bounding boxes (or \"none\" images), unlike YoloV5 and FRCNN which have special ways to process them. I might just use the 0 0 1 1 label but I feel like that might make my model perform worse.\n\nWhat should I do?"
  }
}