{
  "id": 208022,
  "title": " purpose of tain_tfrecords and test_tfrecords? ",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/208022",
  "author_name": "",
  "post_date": "2021-01-01T12:25:07.766566500Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>do we have to use these  records to read images?</p>",
  "messages": [
    {
      "id": "1134607",
      "postDate": "01/01/2021 12:25:07",
      "content": "<p>do we have to use these  records to read images?</p>",
      "rawMarkdown": "do we have to use these  records to read images?",
      "votes": null
    },
    {
      "id": "1134973",
      "postDate": "01/01/2021 18:53:35",
      "content": "<p>You can use either the jpegs or the tfrecords. You don't have to use both.</p>",
      "rawMarkdown": "You can use either the jpegs or the tfrecords. You don't have to use both.",
      "votes": null
    },
    {
      "id": "1135028",
      "postDate": "01/01/2021 20:04:22",
      "content": "<p>Kaggle provides the tfrecords as a favor to us so that those who will be using tensorflow don't have to create the records - they do make running tf keras scripts a lot faster but they are a tiny pain to create correctly.</p>\n<p>When provided they are often not at the full resolution of the jpeg images - in current case the tfrecords are at 512x512 compared to the 800x600 jpegs.</p>",
      "rawMarkdown": "Kaggle provides the tfrecords as a favor to us so that those who will be using tensorflow don't have to create the records - they do make running tf keras scripts a lot faster but they are a tiny pain to create correctly.\n\nWhen provided they are often not at the full resolution of the jpeg images - in current case the tfrecords are at 512x512 compared to the 800x600 jpegs.",
      "votes": null
    },
    {
      "id": "1137551",
      "postDate": "01/04/2021 03:18:34",
      "content": "<p>The TFRecord format is a simple format for storing a sequence of binary records.</p>\n<p>If you are working with large datasets, using a binary file format for storage of your data can have a significant impact on the performance of your import pipeline and as a consequence on the training time of your model. Binary data takes up less space on disk, takes less time to copy and can be read much more efficiently from disk.</p>\n<p>It is very optimized to be used with tensorflow, if that's your cup of tea.<br>\nYou can use use the images also, it doesn't make any difference, the only difference would be the training and read/write times.</p>",
      "rawMarkdown": "The TFRecord format is a simple format for storing a sequence of binary records.\n\nIf you are working with large datasets, using a binary file format for storage of your data can have a significant impact on the performance of your import pipeline and as a consequence on the training time of your model. Binary data takes up less space on disk, takes less time to copy and can be read much more efficiently from disk.\n\nIt is very optimized to be used with tensorflow, if that's your cup of tea.\nYou can use use the images also, it doesn't make any difference, the only difference would be the training and read/write times.",
      "votes": null
    },
    {
      "id": "1154104",
      "postDate": "01/15/2021 12:03:58",
      "content": "<p>what is the purpose of tain_tfrecords and test_tfrecords?</p>",
      "rawMarkdown": "what is the purpose of tain_tfrecords and test_tfrecords?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1134973,
      "author_name": "richardepstein",
      "author_url": "",
      "post_date": "01/01/2021 18:53:35",
      "content": "<p>You can use either the jpegs or the tfrecords. You don't have to use both.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1135028,
      "author_name": "pcjimmmy",
      "author_url": "",
      "post_date": "01/01/2021 20:04:22",
      "content": "<p>Kaggle provides the tfrecords as a favor to us so that those who will be using tensorflow don't have to create the records - they do make running tf keras scripts a lot faster but they are a tiny pain to create correctly.</p>\n<p>When provided they are often not at the full resolution of the jpeg images - in current case the tfrecords are at 512x512 compared to the 800x600 jpegs.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1137551,
      "author_name": "mohneesh7",
      "author_url": "",
      "post_date": "01/04/2021 03:18:34",
      "content": "<p>The TFRecord format is a simple format for storing a sequence of binary records.</p>\n<p>If you are working with large datasets, using a binary file format for storage of your data can have a significant impact on the performance of your import pipeline and as a consequence on the training time of your model. Binary data takes up less space on disk, takes less time to copy and can be read much more efficiently from disk.</p>\n<p>It is very optimized to be used with tensorflow, if that's your cup of tea.<br>\nYou can use use the images also, it doesn't make any difference, the only difference would be the training and read/write times.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1154104,
      "author_name": "sanskarram",
      "author_url": "",
      "post_date": "01/15/2021 12:03:58",
      "content": "<p>what is the purpose of tain_tfrecords and test_tfrecords?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1134607": "do we have to use these  records to read images?",
    "1134973": "You can use either the jpegs or the tfrecords. You don't have to use both.",
    "1135028": "Kaggle provides the tfrecords as a favor to us so that those who will be using tensorflow don't have to create the records - they do make running tf keras scripts a lot faster but they are a tiny pain to create correctly.\n\nWhen provided they are often not at the full resolution of the jpeg images - in current case the tfrecords are at 512x512 compared to the 800x600 jpegs.",
    "1137551": "The TFRecord format is a simple format for storing a sequence of binary records.\n\nIf you are working with large datasets, using a binary file format for storage of your data can have a significant impact on the performance of your import pipeline and as a consequence on the training time of your model. Binary data takes up less space on disk, takes less time to copy and can be read much more efficiently from disk.\n\nIt is very optimized to be used with tensorflow, if that's your cup of tea.\nYou can use use the images also, it doesn't make any difference, the only difference would be the training and read/write times.",
    "1154104": "what is the purpose of tain_tfrecords and test_tfrecords?"
  },
  "source": "meta"
}