{
  "id": 311632,
  "title": "Sidetracked prediction",
  "url": "/competitions/dogs-vs-cats-redux-kernels-edition/discussion/311632",
  "author_name": "",
  "post_date": "2022-03-08T01:34:11.902297400Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi folks!, I would like to know how do you deal with images containing multiple objects, I mean having dogs and cats at the same time, also  I've found some unanimated images, this really makes me think about using Object detection instead so as to draw the bounding boxes around each animal per image. Did you get rid of them? I was thinking avoiding to use them in the training/validation sets as they add bias.</p>\n<p>I look forward to your replies.<br>\nThanks.</p>",
  "messages": [
    {
      "id": "1715421",
      "postDate": "03/08/2022 01:34:11",
      "content": "<p>Hi folks!, I would like to know how do you deal with images containing multiple objects, I mean having dogs and cats at the same time, also  I've found some unanimated images, this really makes me think about using Object detection instead so as to draw the bounding boxes around each animal per image. Did you get rid of them? I was thinking avoiding to use them in the training/validation sets as they add bias.</p>\n<p>I look forward to your replies.<br>\nThanks.</p>",
      "rawMarkdown": "Hi folks!, I would like to know how do you deal with images containing multiple objects, I mean having dogs and cats at the same time, also  I've found some unanimated images, this really makes me think about using Object detection instead so as to draw the bounding boxes around each animal per image. Did you get rid of them? I was thinking avoiding to use them in the training/validation sets as they add bias.\n\nI look forward to your replies.\nThanks.",
      "votes": null
    },
    {
      "id": "2150616",
      "postDate": "02/19/2023 12:54:27",
      "content": "<p>In my opinion relabeling that training image with 0.5 (means 50% dogs and 50% cats) should avoid the bias problem.<br>\nWhat do you think about?<br>\nI was too lazy to use that strategy in my solution, but please let me know if can improve the results.<br>\nEnjoy.</p>",
      "rawMarkdown": "In my opinion relabeling that training image with 0.5 (means 50% dogs and 50% cats) should avoid the bias problem.\nWhat do you think about?\nI was too lazy to use that strategy in my solution, but please let me know if can improve the results.\nEnjoy.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2150616,
      "author_name": "fid0did0",
      "author_url": "",
      "post_date": "02/19/2023 12:54:27",
      "content": "<p>In my opinion relabeling that training image with 0.5 (means 50% dogs and 50% cats) should avoid the bias problem.<br>\nWhat do you think about?<br>\nI was too lazy to use that strategy in my solution, but please let me know if can improve the results.<br>\nEnjoy.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1715421": "Hi folks!, I would like to know how do you deal with images containing multiple objects, I mean having dogs and cats at the same time, also  I've found some unanimated images, this really makes me think about using Object detection instead so as to draw the bounding boxes around each animal per image. Did you get rid of them? I was thinking avoiding to use them in the training/validation sets as they add bias.\n\nI look forward to your replies.\nThanks.",
    "2150616": "In my opinion relabeling that training image with 0.5 (means 50% dogs and 50% cats) should avoid the bias problem.\nWhat do you think about?\nI was too lazy to use that strategy in my solution, but please let me know if can improve the results.\nEnjoy."
  },
  "source": "meta"
}