{
  "id": 186654,
  "title": "ImageDataGenerator",
  "url": "/competitions/landmark-recognition-2020/discussion/186654",
  "author_name": "Tony Ng",
  "post_date": "2020-09-25T12:00:22.966000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I want to find a way to reorganize the files such that ImageDataGenerator from Tensorflow is able to automatically derive the class (landmark_id), but the folder structure currently is not intuitive for that.</p>\n<p>The file size are extremely big, and I am unable to do so in Kaggle given its 4GB storage limit per instance. It is not obvious to me I can create and upload this new structured dataset in Colab as well given its 70GB limit and 429 Error from downloading data.</p>\n<p>Any creative alternatives besides using my very laptop? 😥</p>",
  "messages": [
    {
      "id": 1026545,
      "postDate": "2020-09-25T12:00:22.967Z",
      "content": "<p>I want to find a way to reorganize the files such that ImageDataGenerator from Tensorflow is able to automatically derive the class (landmark_id), but the folder structure currently is not intuitive for that.</p>\n<p>The file size are extremely big, and I am unable to do so in Kaggle given its 4GB storage limit per instance. It is not obvious to me I can create and upload this new structured dataset in Colab as well given its 70GB limit and 429 Error from downloading data.</p>\n<p>Any creative alternatives besides using my very laptop? 😥</p>",
      "rawMarkdown": "I want to find a way to reorganize the files such that ImageDataGenerator from Tensorflow is able to automatically derive the class (landmark_id), but the folder structure currently is not intuitive for that.\n\nThe file size are extremely big, and I am unable to do so in Kaggle given its 4GB storage limit per instance. It is not obvious to me I can create and upload this new structured dataset in Colab as well given its 70GB limit and 429 Error from downloading data.\n\nAny creative alternatives besides using my very laptop? 😥"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1026545": "I want to find a way to reorganize the files such that ImageDataGenerator from Tensorflow is able to automatically derive the class (landmark_id), but the folder structure currently is not intuitive for that.\n\nThe file size are extremely big, and I am unable to do so in Kaggle given its 4GB storage limit per instance. It is not obvious to me I can create and upload this new structured dataset in Colab as well given its 70GB limit and 429 Error from downloading data.\n\nAny creative alternatives besides using my very laptop? 😥"
  }
}