{
  "id": 261011,
  "title": "How to load Faster R-CNN Pytorch without internet?",
  "url": "/competitions/siim-covid19-detection/discussion/261011",
  "author_name": "Jack Ding",
  "post_date": "2021-08-04T02:02:27.292000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I've trained a Faster R-CNN model using the standard torchvision library. Now I want to load said model into the Kaggle notebook for submission. However every time I define the model, it automatically defaults to downloading the resnet backbone from the internet, which I can't do at submission.</p>\n<p>Is there anyway to load Faster R-CNN models without diving into the github repo and reproducing large chunks of code there?</p>",
  "messages": [
    {
      "id": 1443181,
      "postDate": "2021-08-04T02:02:27.293Z",
      "content": "<p>I've trained a Faster R-CNN model using the standard torchvision library. Now I want to load said model into the Kaggle notebook for submission. However every time I define the model, it automatically defaults to downloading the resnet backbone from the internet, which I can't do at submission.</p>\n<p>Is there anyway to load Faster R-CNN models without diving into the github repo and reproducing large chunks of code there?</p>",
      "rawMarkdown": "I've trained a Faster R-CNN model using the standard torchvision library. Now I want to load said model into the Kaggle notebook for submission. However every time I define the model, it automatically defaults to downloading the resnet backbone from the internet, which I can't do at submission.\n\nIs there anyway to load Faster R-CNN models without diving into the github repo and reproducing large chunks of code there?"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1443181": "I've trained a Faster R-CNN model using the standard torchvision library. Now I want to load said model into the Kaggle notebook for submission. However every time I define the model, it automatically defaults to downloading the resnet backbone from the internet, which I can't do at submission.\n\nIs there anyway to load Faster R-CNN models without diving into the github repo and reproducing large chunks of code there?"
  }
}