{
  "id": 206610,
  "title": "Size of binary encoded images seems to be smaller than jpgs",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/206610",
  "author_name": "",
  "post_date": "2020-12-25T13:51:48.977316100Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>is it intended that the size of the images contained in the tfrecord files is smaller (512x512) than the original jpg images (800x600) or did I miss something?`</p>\n<p>Thanks and best regards</p>",
  "messages": [
    {
      "id": "1126296",
      "postDate": "12/25/2020 13:51:48",
      "content": "<p>Hi,</p>\n<p>is it intended that the size of the images contained in the tfrecord files is smaller (512x512) than the original jpg images (800x600) or did I miss something?`</p>\n<p>Thanks and best regards</p>",
      "rawMarkdown": "Hi,\n\nis it intended that the size of the images contained in the tfrecord files is smaller (512x512) than the original jpg images (800x600) or did I miss something?`\n\n\nThanks and best regards",
      "votes": null
    },
    {
      "id": "1127000",
      "postDate": "12/26/2020 06:49:43",
      "content": "<p>Many kernels in this competition choose to use (512x512) size image because many model like efficientnet use 512x512 size images. So many kernels crop or cut original images to 512x512. Of course you can build your own CNN model using 800x600 size images.</p>",
      "rawMarkdown": "Many kernels in this competition choose to use (512x512) size image because many model like efficientnet use 512x512 size images. So many kernels crop or cut original images to 512x512. Of course you can build your own CNN model using 800x600 size images.",
      "votes": null
    },
    {
      "id": "1127938",
      "postDate": "12/27/2020 03:59:32",
      "content": "<p>Kaggle normaly lets us know the size of the tfrecord files in the overview or data tabs - the files are provided for convenience and are almost NEVER in the same image size as the jpgs, but rather are in a size(s) that most of us would use if creating them yourself.</p>",
      "rawMarkdown": "Kaggle normaly lets us know the size of the tfrecord files in the overview or data tabs - the files are provided for convenience and are almost NEVER in the same image size as the jpgs, but rather are in a size(s) that most of us would use if creating them yourself.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1127000,
      "author_name": "vkehfdl1",
      "author_url": "",
      "post_date": "12/26/2020 06:49:43",
      "content": "<p>Many kernels in this competition choose to use (512x512) size image because many model like efficientnet use 512x512 size images. So many kernels crop or cut original images to 512x512. Of course you can build your own CNN model using 800x600 size images.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1127938,
      "author_name": "pcjimmmy",
      "author_url": "",
      "post_date": "12/27/2020 03:59:32",
      "content": "<p>Kaggle normaly lets us know the size of the tfrecord files in the overview or data tabs - the files are provided for convenience and are almost NEVER in the same image size as the jpgs, but rather are in a size(s) that most of us would use if creating them yourself.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1126296": "Hi,\n\nis it intended that the size of the images contained in the tfrecord files is smaller (512x512) than the original jpg images (800x600) or did I miss something?`\n\n\nThanks and best regards",
    "1127000": "Many kernels in this competition choose to use (512x512) size image because many model like efficientnet use 512x512 size images. So many kernels crop or cut original images to 512x512. Of course you can build your own CNN model using 800x600 size images.",
    "1127938": "Kaggle normaly lets us know the size of the tfrecord files in the overview or data tabs - the files are provided for convenience and are almost NEVER in the same image size as the jpgs, but rather are in a size(s) that most of us would use if creating them yourself."
  },
  "source": "meta"
}