{
  "id": 177059,
  "title": "Dataset Transformation Pipeline",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/177059",
  "author_name": "",
  "post_date": "2020-08-24T16:14:30.058032Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I have been considering creating a flexible transformation pipeline to convert the Kaggle provided datasets into ConvNet training datasets, but before I do that I thought It was worth asking if there is already an easy way to transform variable datasets like videos and large images into smaller 224 x 224 tiles saved as .png? </p>\n<p>There has to be an easier way than writing a new custom transformation for each competition? </p>",
  "messages": [
    {
      "id": "983833",
      "postDate": "08/24/2020 16:14:30",
      "content": "<p>I have been considering creating a flexible transformation pipeline to convert the Kaggle provided datasets into ConvNet training datasets, but before I do that I thought It was worth asking if there is already an easy way to transform variable datasets like videos and large images into smaller 224 x 224 tiles saved as .png? </p>\n<p>There has to be an easier way than writing a new custom transformation for each competition? </p>",
      "rawMarkdown": "I have been considering creating a flexible transformation pipeline to convert the Kaggle provided datasets into ConvNet training datasets, but before I do that I thought It was worth asking if there is already an easy way to transform variable datasets like videos and large images into smaller 224 x 224 tiles saved as .png? \n\nThere has to be an easier way than writing a new custom transformation for each competition?",
      "votes": null
    },
    {
      "id": "984276",
      "postDate": "08/25/2020 02:30:48",
      "content": "<p>Hi! <a href=\"https://www.kaggle.com/jonmacpherson\" target=\"_blank\">@jonmacpherson</a> This Course has really good tips on how to rely on Tensorflow built-in libraries for this:<br>\n<a href=\"url\" target=\"_blank\">https://www.coursera.org/learn/convolutional-neural-networks-tensorflow</a></p>",
      "rawMarkdown": "Hi! @jonmacpherson This Course has really good tips on how to rely on Tensorflow built-in libraries for this:\n[https://www.coursera.org/learn/convolutional-neural-networks-tensorflow](url)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 984276,
      "author_name": "cv13j0",
      "author_url": "",
      "post_date": "08/25/2020 02:30:48",
      "content": "<p>Hi! <a href=\"https://www.kaggle.com/jonmacpherson\" target=\"_blank\">@jonmacpherson</a> This Course has really good tips on how to rely on Tensorflow built-in libraries for this:<br>\n<a href=\"url\" target=\"_blank\">https://www.coursera.org/learn/convolutional-neural-networks-tensorflow</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "983833": "I have been considering creating a flexible transformation pipeline to convert the Kaggle provided datasets into ConvNet training datasets, but before I do that I thought It was worth asking if there is already an easy way to transform variable datasets like videos and large images into smaller 224 x 224 tiles saved as .png? \n\nThere has to be an easier way than writing a new custom transformation for each competition?",
    "984276": "Hi! @jonmacpherson This Course has really good tips on how to rely on Tensorflow built-in libraries for this:\n[https://www.coursera.org/learn/convolutional-neural-networks-tensorflow](url)"
  },
  "source": "meta"
}