{
  "id": 168208,
  "title": "I made a dataset for PyTorch users that loads the jpg images easier!",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/168208",
  "author_name": "",
  "post_date": "2020-07-19T16:16:10.978929100Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I downloaded the provided data set for the SIIM-ISIC Melanoma Classification Competition and elected to use the JPEG images. To load the training and testing images into a PyTorch ImageFolder the images must be sorted into subfolders of their respective classifications. That's where this data set came from. \n<a href=\"https://www.kaggle.com/whatsthevariance/melanoma-image-data\">https://www.kaggle.com/whatsthevariance/melanoma-image-data</a></p>\n\n<p>The function I wrote to sort the images is in my kernel:\n<a href=\"https://www.kaggle.com/whatsthevariance/diagnosing-skin-cancer-with-bagged-neural-nets\">https://www.kaggle.com/whatsthevariance/diagnosing-skin-cancer-with-bagged-neural-nets</a>\nshould you wish to duplicate this locally.</p>",
  "messages": [
    {
      "id": "935776",
      "postDate": "07/19/2020 16:16:10",
      "content": "<p>I downloaded the provided data set for the SIIM-ISIC Melanoma Classification Competition and elected to use the JPEG images. To load the training and testing images into a PyTorch ImageFolder the images must be sorted into subfolders of their respective classifications. That's where this data set came from. \n<a href=\"https://www.kaggle.com/whatsthevariance/melanoma-image-data\">https://www.kaggle.com/whatsthevariance/melanoma-image-data</a></p>\n\n<p>The function I wrote to sort the images is in my kernel:\n<a href=\"https://www.kaggle.com/whatsthevariance/diagnosing-skin-cancer-with-bagged-neural-nets\">https://www.kaggle.com/whatsthevariance/diagnosing-skin-cancer-with-bagged-neural-nets</a>\nshould you wish to duplicate this locally.</p>",
      "rawMarkdown": "I downloaded the provided data set for the SIIM-ISIC Melanoma Classification Competition and elected to use the JPEG images. To load the training and testing images into a PyTorch ImageFolder the images must be sorted into subfolders of their respective classifications. That's where this data set came from. \nhttps://www.kaggle.com/whatsthevariance/melanoma-image-data\n\nThe function I wrote to sort the images is in my kernel:\nhttps://www.kaggle.com/whatsthevariance/diagnosing-skin-cancer-with-bagged-neural-nets\nshould you wish to duplicate this locally.",
      "votes": null
    },
    {
      "id": "935908",
      "postDate": "07/19/2020 19:08:05",
      "content": "<p>what size jpegs have you kept here?\nThe original sizes?</p>",
      "rawMarkdown": "what size jpegs have you kept here?\nThe original sizes?",
      "votes": null
    },
    {
      "id": "936017",
      "postDate": "07/19/2020 22:24:48",
      "content": "<p>yup, original sizes</p>",
      "rawMarkdown": "yup, original sizes",
      "votes": null
    },
    {
      "id": "936038",
      "postDate": "07/19/2020 23:09:14",
      "content": "<p>In your solution images are resized on the fly, right? Do you have any benchmarks how much slower it is from jpegs which are already resized?</p>",
      "rawMarkdown": "In your solution images are resized on the fly, right? Do you have any benchmarks how much slower it is from jpegs which are already resized?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 935908,
      "author_name": "sayakdasgupta",
      "author_url": "",
      "post_date": "07/19/2020 19:08:05",
      "content": "<p>what size jpegs have you kept here?\nThe original sizes?</p>",
      "votes": null,
      "replies": [
        {
          "id": 936017,
          "author_name": "whatsthevariance",
          "author_url": "",
          "post_date": "07/19/2020 22:24:48",
          "content": "<p>yup, original sizes</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 936038,
      "author_name": "jacekpoplawski",
      "author_url": "",
      "post_date": "07/19/2020 23:09:14",
      "content": "<p>In your solution images are resized on the fly, right? Do you have any benchmarks how much slower it is from jpegs which are already resized?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "935776": "I downloaded the provided data set for the SIIM-ISIC Melanoma Classification Competition and elected to use the JPEG images. To load the training and testing images into a PyTorch ImageFolder the images must be sorted into subfolders of their respective classifications. That's where this data set came from. \nhttps://www.kaggle.com/whatsthevariance/melanoma-image-data\n\nThe function I wrote to sort the images is in my kernel:\nhttps://www.kaggle.com/whatsthevariance/diagnosing-skin-cancer-with-bagged-neural-nets\nshould you wish to duplicate this locally.",
    "935908": "what size jpegs have you kept here?\nThe original sizes?",
    "936017": "yup, original sizes",
    "936038": "In your solution images are resized on the fly, right? Do you have any benchmarks how much slower it is from jpegs which are already resized?"
  },
  "source": "meta"
}