{
  "id": 166605,
  "title": "Resize the image from tfrecords",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/166605",
  "author_name": "",
  "post_date": "2020-07-13T13:35:34.536812600Z",
  "votes": 2,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I am first time working with tfrecords and new to Computer vision.</p>\n\n<p>So please provide valuable feedback and suggestions.   </p>\n\n<p>Step 1- Read the train tfrecord \nStep 2 - Parse an image and label from the given serialized_example (done for all the examples)\nTotal images = 2071,Total labels = 2071</p>\n\n<p>Length of first image (258372,)\nLength of second image (264912,)\nso on...</p>\n\n<p>All the image is of different size.Can someone please help me to understand how we can resize all the image to same shape.</p>",
  "messages": [
    {
      "id": "927562",
      "postDate": "07/13/2020 13:35:34",
      "content": "<p>I am first time working with tfrecords and new to Computer vision.</p>\n\n<p>So please provide valuable feedback and suggestions.   </p>\n\n<p>Step 1- Read the train tfrecord \nStep 2 - Parse an image and label from the given serialized_example (done for all the examples)\nTotal images = 2071,Total labels = 2071</p>\n\n<p>Length of first image (258372,)\nLength of second image (264912,)\nso on...</p>\n\n<p>All the image is of different size.Can someone please help me to understand how we can resize all the image to same shape.</p>",
      "rawMarkdown": "I am first time working with tfrecords and new to Computer vision.\n\nSo please provide valuable feedback and suggestions.   \n\nStep 1- Read the train tfrecord \nStep 2 - Parse an image and label from the given serialized_example (done for all the examples)\nTotal images = 2071,Total labels = 2071\n\nLength of first image (258372,)\nLength of second image (264912,)\nso on...\n\nAll the image is of different size.Can someone please help me to understand how we can resize all the image to same shape.",
      "votes": null
    },
    {
      "id": "927822",
      "postDate": "07/13/2020 15:42:11",
      "content": "<p>IMHO, the best way to do it is to make a new set of tfrecords resized to the desired size. Here is an example illustrating how this can be done in TensorFlow:\n<a href=\"https://www.kaggle.com/graf10a/siim-example-of-making-tfrec-files-512x512\">SIIM Example of making tfrec files 512x512</a></p>\n\n<p>One alternative would be to include the resize function into your training pipeline but I would not recommend that -- it might slow down your training significantly. </p>",
      "rawMarkdown": "IMHO, the best way to do it is to make a new set of tfrecords resized to the desired size. Here is an example illustrating how this can be done in TensorFlow:\n[SIIM Example of making tfrec files 512x512](https://www.kaggle.com/graf10a/siim-example-of-making-tfrec-files-512x512)\n\nOne alternative would be to include the resize function into your training pipeline but I would not recommend that -- it might slow down your training significantly.",
      "votes": null
    },
    {
      "id": "928013",
      "postDate": "07/13/2020 17:28:12",
      "content": "<p>Thank you very much for quick help.</p>",
      "rawMarkdown": "Thank you very much for quick help.",
      "votes": null
    },
    {
      "id": "928211",
      "postDate": "07/13/2020 19:47:49",
      "content": "<p>Try to add padding. Like NLP🙌 </p>",
      "rawMarkdown": "Try to add padding. Like NLP🙌",
      "votes": null
    },
    {
      "id": "928482",
      "postDate": "07/14/2020 03:18:44",
      "content": "<p>Thanks Sure Will try that.</p>",
      "rawMarkdown": "Thanks Sure Will try that.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 927822,
      "author_name": "graf10a",
      "author_url": "",
      "post_date": "07/13/2020 15:42:11",
      "content": "<p>IMHO, the best way to do it is to make a new set of tfrecords resized to the desired size. Here is an example illustrating how this can be done in TensorFlow:\n<a href=\"https://www.kaggle.com/graf10a/siim-example-of-making-tfrec-files-512x512\">SIIM Example of making tfrec files 512x512</a></p>\n\n<p>One alternative would be to include the resize function into your training pipeline but I would not recommend that -- it might slow down your training significantly. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 928013,
      "author_name": "prince711",
      "author_url": "",
      "post_date": "07/13/2020 17:28:12",
      "content": "<p>Thank you very much for quick help.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 928211,
      "author_name": "doanquanvietnamca",
      "author_url": "",
      "post_date": "07/13/2020 19:47:49",
      "content": "<p>Try to add padding. Like NLP🙌 </p>",
      "votes": null,
      "replies": [
        {
          "id": 928482,
          "author_name": "prince711",
          "author_url": "",
          "post_date": "07/14/2020 03:18:44",
          "content": "<p>Thanks Sure Will try that.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "927562": "I am first time working with tfrecords and new to Computer vision.\n\nSo please provide valuable feedback and suggestions.   \n\nStep 1- Read the train tfrecord \nStep 2 - Parse an image and label from the given serialized_example (done for all the examples)\nTotal images = 2071,Total labels = 2071\n\nLength of first image (258372,)\nLength of second image (264912,)\nso on...\n\nAll the image is of different size.Can someone please help me to understand how we can resize all the image to same shape.",
    "927822": "IMHO, the best way to do it is to make a new set of tfrecords resized to the desired size. Here is an example illustrating how this can be done in TensorFlow:\n[SIIM Example of making tfrec files 512x512](https://www.kaggle.com/graf10a/siim-example-of-making-tfrec-files-512x512)\n\nOne alternative would be to include the resize function into your training pipeline but I would not recommend that -- it might slow down your training significantly.",
    "928013": "Thank you very much for quick help.",
    "928211": "Try to add padding. Like NLP🙌",
    "928482": "Thanks Sure Will try that."
  },
  "source": "meta"
}