{
  "id": 153191,
  "title": "How to deal with images with no object? (Apart form removing them)",
  "url": "/competitions/global-wheat-detection/discussion/153191",
  "author_name": "",
  "post_date": "2020-05-23T15:06:29.101939700Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>How to make use of images with no objects?</p>",
  "messages": [
    {
      "id": "858536",
      "postDate": "05/23/2020 15:06:29",
      "content": "<p>How to make use of images with no objects?</p>",
      "rawMarkdown": "How to make use of images with no objects?",
      "votes": null
    },
    {
      "id": "858710",
      "postDate": "05/23/2020 18:12:05",
      "content": "<p>well we can train a network with object or no object then combine it with object detection model can we ?</p>",
      "rawMarkdown": "well we can train a network with object or no object then combine it with object detection model can we ?",
      "votes": null
    },
    {
      "id": "858802",
      "postDate": "05/23/2020 20:17:23",
      "content": "<p>It depends on what your implementation is of course. In torchvision FasterRCNN, just pass torch.zeros((0,4)) as the boxes </p>",
      "rawMarkdown": "It depends on what your implementation is of course. In torchvision FasterRCNN, just pass torch.zeros((0,4)) as the boxes",
      "votes": null
    },
    {
      "id": "859439",
      "postDate": "05/24/2020 12:29:20",
      "content": "<p>ohh, nice! thank you.</p>",
      "rawMarkdown": "ohh, nice! thank you.",
      "votes": null
    },
    {
      "id": "859442",
      "postDate": "05/24/2020 12:29:51",
      "content": "<p>Thanks, I also thought of similar thing!</p>",
      "rawMarkdown": "Thanks, I also thought of similar thing!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 858710,
      "author_name": "ajax0564",
      "author_url": "",
      "post_date": "05/23/2020 18:12:05",
      "content": "<p>well we can train a network with object or no object then combine it with object detection model can we ?</p>",
      "votes": null,
      "replies": [
        {
          "id": 859442,
          "author_name": "jimitshah777",
          "author_url": "",
          "post_date": "05/24/2020 12:29:51",
          "content": "<p>Thanks, I also thought of similar thing!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 858802,
      "author_name": "matthewmasters",
      "author_url": "",
      "post_date": "05/23/2020 20:17:23",
      "content": "<p>It depends on what your implementation is of course. In torchvision FasterRCNN, just pass torch.zeros((0,4)) as the boxes </p>",
      "votes": null,
      "replies": [
        {
          "id": 859439,
          "author_name": "jimitshah777",
          "author_url": "",
          "post_date": "05/24/2020 12:29:20",
          "content": "<p>ohh, nice! thank you.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "858536": "How to make use of images with no objects?",
    "858710": "well we can train a network with object or no object then combine it with object detection model can we ?",
    "858802": "It depends on what your implementation is of course. In torchvision FasterRCNN, just pass torch.zeros((0,4)) as the boxes",
    "859439": "ohh, nice! thank you.",
    "859442": "Thanks, I also thought of similar thing!"
  },
  "source": "meta"
}