{
  "id": 263668,
  "title": "EfficientDet - Out of curiousity",
  "url": "/competitions/siim-covid19-detection/discussion/263668",
  "author_name": "Michael",
  "post_date": "2021-08-10T01:30:20.324000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi Everyone,</p>\n<p>Thanks for a great competition and congratulations to the winners. </p>\n<p>I was just wondering whether anyone used a modification of my public kernels <a href=\"https://www.kaggle.com/mikeb127/efficientdet-pytorch-training-starter\" target=\"_blank\">here</a> to train an EfficientDet model for their final submission, and if they did how it went? I trained a rather vanilla D5 which seemed to work ok (.592 private), but expecting that you probably needed to use extensive data augmentation and model ensembling to do well…… </p>\n<p>Once again, thanks to everybody for participating and making it a great competition</p>",
  "messages": [
    {
      "id": 1462719,
      "postDate": "2021-08-10T01:30:20.323Z",
      "content": "<p>Hi Everyone,</p>\n<p>Thanks for a great competition and congratulations to the winners. </p>\n<p>I was just wondering whether anyone used a modification of my public kernels <a href=\"https://www.kaggle.com/mikeb127/efficientdet-pytorch-training-starter\" target=\"_blank\">here</a> to train an EfficientDet model for their final submission, and if they did how it went? I trained a rather vanilla D5 which seemed to work ok (.592 private), but expecting that you probably needed to use extensive data augmentation and model ensembling to do well…… </p>\n<p>Once again, thanks to everybody for participating and making it a great competition</p>",
      "rawMarkdown": "Hi Everyone,\n\nThanks for a great competition and congratulations to the winners. \n\nI was just wondering whether anyone used a modification of my public kernels [here](https://www.kaggle.com/mikeb127/efficientdet-pytorch-training-starter) to train an EfficientDet model for their final submission, and if they did how it went? I trained a rather vanilla D5 which seemed to work ok (.592 private), but expecting that you probably needed to use extensive data augmentation and model ensembling to do well...... \n\nOnce again, thanks to everybody for participating and making it a great competition",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1462719": "Hi Everyone,\n\nThanks for a great competition and congratulations to the winners. \n\nI was just wondering whether anyone used a modification of my public kernels [here](https://www.kaggle.com/mikeb127/efficientdet-pytorch-training-starter) to train an EfficientDet model for their final submission, and if they did how it went? I trained a rather vanilla D5 which seemed to work ok (.592 private), but expecting that you probably needed to use extensive data augmentation and model ensembling to do well...... \n\nOnce again, thanks to everybody for participating and making it a great competition"
  }
}