{
  "id": 311438,
  "title": "Proper way to get resized datasets",
  "url": "/competitions/happy-whale-and-dolphin/discussion/311438",
  "author_name": "Timothee Wright",
  "post_date": "2022-03-06T22:36:02.790000",
  "votes": 11,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hey everyone,<br>\nI'm still a beginner in computer vision so maybe I'm mixing several concepts.</p>\n<p>I saw that <a href=\"https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/304686\" target=\"_blank\">RDizzl3 proposed a reduced data</a> using bicubic interpolation.</p>\n<p>I'm wandering if it would be best to first use bounding boxes and crop the interesting subject and then resize rather than resizing the image directly in order not to destroy information about the object.</p>\n<p>Also could squeezing an image be a problem for CNNs ? Is it better to do some padding to get square images ?</p>\n<p>Any thoughts could be useful for this challenge and for later :) </p>\n<p>Thank you !</p>",
  "messages": [
    {
      "id": 1714374,
      "postDate": "2022-03-06T22:36:02.790Z",
      "content": "<p>Hey everyone,<br>\nI'm still a beginner in computer vision so maybe I'm mixing several concepts.</p>\n<p>I saw that <a href=\"https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/304686\" target=\"_blank\">RDizzl3 proposed a reduced data</a> using bicubic interpolation.</p>\n<p>I'm wandering if it would be best to first use bounding boxes and crop the interesting subject and then resize rather than resizing the image directly in order not to destroy information about the object.</p>\n<p>Also could squeezing an image be a problem for CNNs ? Is it better to do some padding to get square images ?</p>\n<p>Any thoughts could be useful for this challenge and for later :) </p>\n<p>Thank you !</p>",
      "rawMarkdown": "Hey everyone,\nI'm still a beginner in computer vision so maybe I'm mixing several concepts.\n\nI saw that [RDizzl3 proposed a reduced data](https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/304686) using bicubic interpolation.\n\nI'm wandering if it would be best to first use bounding boxes and crop the interesting subject and then resize rather than resizing the image directly in order not to destroy information about the object.\n\nAlso could squeezing an image be a problem for CNNs ? Is it better to do some padding to get square images ?\n\nAny thoughts could be useful for this challenge and for later :) \n\nThank you !",
      "votes": 11
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1714374": "Hey everyone,\nI'm still a beginner in computer vision so maybe I'm mixing several concepts.\n\nI saw that [RDizzl3 proposed a reduced data](https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/304686) using bicubic interpolation.\n\nI'm wandering if it would be best to first use bounding boxes and crop the interesting subject and then resize rather than resizing the image directly in order not to destroy information about the object.\n\nAlso could squeezing an image be a problem for CNNs ? Is it better to do some padding to get square images ?\n\nAny thoughts could be useful for this challenge and for later :) \n\nThank you !"
  }
}