{
  "id": 200807,
  "title": "CenterCrop + Resize on validation dataset",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/200807",
  "author_name": "",
  "post_date": "2020-12-02T00:20:12.667661Z",
  "votes": 4,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Across many Pytorch notebooks i can see the same pair of transformations, applied to a validation dataset:</p>\n<pre><code>albumentations.CenterCrop(IMG_SIZE, IMG_SIZE, p=1.),\nalbumentations.Resize(IMG_SIZE, IMG_SIZE),\n</code></pre>\n<p>With CenterCrop an image is cropped to the size of IMG_SIZE x IMG_SIZE already. What's the sense of Resize transformation here?</p>",
  "messages": [
    {
      "id": "1098921",
      "postDate": "12/02/2020 00:20:12",
      "content": "<p>Across many Pytorch notebooks i can see the same pair of transformations, applied to a validation dataset:</p>\n<pre><code>albumentations.CenterCrop(IMG_SIZE, IMG_SIZE, p=1.),\nalbumentations.Resize(IMG_SIZE, IMG_SIZE),\n</code></pre>\n<p>With CenterCrop an image is cropped to the size of IMG_SIZE x IMG_SIZE already. What's the sense of Resize transformation here?</p>",
      "rawMarkdown": "Across many Pytorch notebooks i can see the same pair of transformations, applied to a validation dataset:\n\n```\nalbumentations.CenterCrop(IMG_SIZE, IMG_SIZE, p=1.),\nalbumentations.Resize(IMG_SIZE, IMG_SIZE),\n```\nWith CenterCrop an image is cropped to the size of IMG_SIZE x IMG_SIZE already. What's the sense of Resize transformation here?",
      "votes": null
    },
    {
      "id": "1098963",
      "postDate": "12/02/2020 01:28:20",
      "content": "<p>I also couldn't make sense out of it :)</p>",
      "rawMarkdown": "I also couldn't make sense out of it :)",
      "votes": null
    },
    {
      "id": "1098992",
      "postDate": "12/02/2020 02:12:29",
      "content": "<p>Great! It means I have my doubts but I'm not alone in this and finally can go to sleep.</p>",
      "rawMarkdown": "Great! It means I have my doubts but I'm not alone in this and finally can go to sleep.",
      "votes": null
    },
    {
      "id": "1099019",
      "postDate": "12/02/2020 02:56:49",
      "content": "<p>True, I am confused now</p>",
      "rawMarkdown": "True, I am confused now",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1098963,
      "author_name": "keremt",
      "author_url": "",
      "post_date": "12/02/2020 01:28:20",
      "content": "<p>I also couldn't make sense out of it :)</p>",
      "votes": null,
      "replies": [
        {
          "id": 1098992,
          "author_name": "dunklerwald",
          "author_url": "",
          "post_date": "12/02/2020 02:12:29",
          "content": "<p>Great! It means I have my doubts but I'm not alone in this and finally can go to sleep.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1099019,
          "author_name": "reighns",
          "author_url": "",
          "post_date": "12/02/2020 02:56:49",
          "content": "<p>True, I am confused now</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1098921": "Across many Pytorch notebooks i can see the same pair of transformations, applied to a validation dataset:\n\n```\nalbumentations.CenterCrop(IMG_SIZE, IMG_SIZE, p=1.),\nalbumentations.Resize(IMG_SIZE, IMG_SIZE),\n```\nWith CenterCrop an image is cropped to the size of IMG_SIZE x IMG_SIZE already. What's the sense of Resize transformation here?",
    "1098963": "I also couldn't make sense out of it :)",
    "1098992": "Great! It means I have my doubts but I'm not alone in this and finally can go to sleep.",
    "1099019": "True, I am confused now"
  },
  "source": "meta"
}