{
  "id": 421010,
  "title": "How to resize images in segmentation ?",
  "url": "/competitions/google-research-identify-contrails-reduce-global-warming/discussion/421010",
  "author_name": "k_tomo",
  "post_date": "2023-07-03T15:07:02.344000",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>This is first time to try segmentation  task. <br>\nAt first step, I want to resize images to improve evaluation.<br>\nIs there anything I should care?<br>\n(Maybe, should I chage mask image size, too?)</p>",
  "messages": [
    {
      "id": 2328468,
      "postDate": "2023-07-03T15:41:57.317Z",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/whitelily\" target=\"_blank\">@whitelily</a>,<br>\nYou can use Albumentation's Resize to resize your images to the desired size. Don't resize the masks, just use F.interpolate to resize your model's predictions back to 256by256 just before giving them to the loss.<br>\nHope this helps.</p>",
      "rawMarkdown": "Hi @whitelily,\nYou can use Albumentation's Resize to resize your images to the desired size. Don't resize the masks, just use F.interpolate to resize your model's predictions back to 256by256 just before giving them to the loss.\nHope this helps.",
      "votes": 4,
      "replies": [
        {
          "id": 2328494,
          "postDate": "2023-07-03T16:17:21.183Z",
          "content": "<p>Thank you for your kind and quick advice !!! <br>\nI'll try it !</p>",
          "rawMarkdown": "Thank you for your kind and quick advice !!! \nI'll try it !",
          "votes": 1,
          "replies": [
            {
              "id": 2334054,
              "postDate": "2023-07-07T12:10:58.673Z",
              "content": "<p>Thanks great advice <a href=\"https://www.kaggle.com/shashwatraman\" target=\"_blank\">@shashwatraman</a> 👍 Think <a href=\"https://www.kaggle.com/whitelily\" target=\"_blank\">@whitelily</a> you can refer to some public repos that do this already .</p>",
              "rawMarkdown": "Thanks great advice @shashwatraman 👍 Think @whitelily you can refer to some public repos that do this already ."
            }
          ]
        }
      ]
    },
    {
      "id": 2328405,
      "postDate": "2023-07-03T15:07:02.343Z",
      "content": "<p>This is first time to try segmentation  task. <br>\nAt first step, I want to resize images to improve evaluation.<br>\nIs there anything I should care?<br>\n(Maybe, should I chage mask image size, too?)</p>",
      "rawMarkdown": "This is first time to try segmentation  task. \nAt first step, I want to resize images to improve evaluation.\nIs there anything I should care?\n(Maybe, should I chage mask image size, too?)",
      "votes": 2
    }
  ],
  "comments": [
    {
      "id": 2328468,
      "author_name": "Shashwat Raman",
      "author_url": "",
      "post_date": "2023-07-03T15:41:57.317000",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/whitelily\" target=\"_blank\">@whitelily</a>,<br>\nYou can use Albumentation's Resize to resize your images to the desired size. Don't resize the masks, just use F.interpolate to resize your model's predictions back to 256by256 just before giving them to the loss.<br>\nHope this helps.</p>",
      "votes": 4,
      "replies": [
        {
          "id": 2328494,
          "author_name": "k_tomo",
          "author_url": "",
          "post_date": "2023-07-03T16:17:21.183000",
          "content": "<p>Thank you for your kind and quick advice !!! <br>\nI'll try it !</p>",
          "votes": 1,
          "replies": [
            {
              "id": 2334054,
              "author_name": "Gaurav Rawat",
              "author_url": "",
              "post_date": "2023-07-07T12:10:58.673000",
              "content": "<p>Thanks great advice <a href=\"https://www.kaggle.com/shashwatraman\" target=\"_blank\">@shashwatraman</a> 👍 Think <a href=\"https://www.kaggle.com/whitelily\" target=\"_blank\">@whitelily</a> you can refer to some public repos that do this already .</p>",
              "votes": 0,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2328468": "Hi @whitelily,\nYou can use Albumentation's Resize to resize your images to the desired size. Don't resize the masks, just use F.interpolate to resize your model's predictions back to 256by256 just before giving them to the loss.\nHope this helps.",
    "2328405": "This is first time to try segmentation  task. \nAt first step, I want to resize images to improve evaluation.\nIs there anything I should care?\n(Maybe, should I chage mask image size, too?)"
  }
}