{
  "id": 354105,
  "title": "how to ensemble Bounding Boxes from different models",
  "url": "/competitions/rsna-2022-cervical-spine-fracture-detection/discussion/354105",
  "author_name": "Altair Farooque",
  "post_date": "2022-09-21T06:09:31.558000",
  "votes": 5,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi , i'm Altair .</p>\n<p>I have build a custom object detection model , this model will ouptuts 3 stage Bboxes .<br>\nI wanna ensemble all B-boxes from 3 stages. But the model will output fixed no. of B-boxes at each stage .</p>\n<p>I have tried some methods like Weighted- Box - fusion , etc . but this didn't work , since class labels are always <br>\nless than no. of bounding boxes .</p>\n<p>it would be great to know is there other methods do this.</p>",
  "messages": [
    {
      "id": 1948515,
      "postDate": "2022-09-21T06:09:31.557Z",
      "content": "<p>Hi , i'm Altair .</p>\n<p>I have build a custom object detection model , this model will ouptuts 3 stage Bboxes .<br>\nI wanna ensemble all B-boxes from 3 stages. But the model will output fixed no. of B-boxes at each stage .</p>\n<p>I have tried some methods like Weighted- Box - fusion , etc . but this didn't work , since class labels are always <br>\nless than no. of bounding boxes .</p>\n<p>it would be great to know is there other methods do this.</p>",
      "rawMarkdown": "Hi , i'm Altair .\n\nI have build a custom object detection model , this model will ouptuts 3 stage Bboxes .\nI wanna ensemble all B-boxes from 3 stages. But the model will output fixed no. of B-boxes at each stage .\n\nI have tried some methods like Weighted- Box - fusion , etc . but this didn't work , since class labels are always \nless than no. of bounding boxes .\n\nit would be great to know is there other methods do this.",
      "votes": 5
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1948515": "Hi , i'm Altair .\n\nI have build a custom object detection model , this model will ouptuts 3 stage Bboxes .\nI wanna ensemble all B-boxes from 3 stages. But the model will output fixed no. of B-boxes at each stage .\n\nI have tried some methods like Weighted- Box - fusion , etc . but this didn't work , since class labels are always \nless than no. of bounding boxes .\n\nit would be great to know is there other methods do this."
  }
}