{
  "id": 240512,
  "title": "What's the size of the hidden test set?",
  "url": "/competitions/siim-covid19-detection/discussion/240512",
  "author_name": "",
  "post_date": "2021-05-20T06:23:28.424375200Z",
  "votes": 2,
  "comment_count": 7,
  "views": 0,
  "content": "<p>I don't see it mentioned anywhere; or am I just missing something obvious?<br>\nThanks</p>",
  "messages": [
    {
      "id": "1315841",
      "postDate": "05/20/2021 06:23:28",
      "content": "<p>I don't see it mentioned anywhere; or am I just missing something obvious?<br>\nThanks</p>",
      "rawMarkdown": "I don't see it mentioned anywhere; or am I just missing something obvious?\nThanks",
      "votes": null
    },
    {
      "id": "1316178",
      "postDate": "05/20/2021 11:02:06",
      "content": "<p>It's mentioned in <a href=\"https://www.kaggle.com/c/siim-covid19-detection/data\" target=\"_blank\">Data</a>.<br>\n<code>The hidden test dataset is of roughly the same scale as the training dataset.</code></p>",
      "rawMarkdown": "It's mentioned in [Data](https://www.kaggle.com/c/siim-covid19-detection/data).\n`The hidden test dataset is of roughly the same scale as the training dataset.`",
      "votes": null
    },
    {
      "id": "1327710",
      "postDate": "05/29/2021 14:57:25",
      "content": "<p>The images that are in the \"test\" folder. Are they from the public part of the test dataset?</p>",
      "rawMarkdown": "The images that are in the \"test\" folder. Are they from the public part of the test dataset?",
      "votes": null
    },
    {
      "id": "1327803",
      "postDate": "05/29/2021 16:26:22",
      "content": "<p>All Images are in the test folder, otherwise how could you submit solution? Images are in dicom format but most people convert them into jpg or png so If there are any more Images than in test folder they won’t be converted and couldn’t be used in the kernel.</p>",
      "rawMarkdown": "All Images are in the test folder, otherwise how could you submit solution? Images are in dicom format but most people convert them into jpg or png so If there are any more Images than in test folder they won’t be converted and couldn’t be used in the kernel.",
      "votes": null
    },
    {
      "id": "1327808",
      "postDate": "05/29/2021 16:28:15",
      "content": "<p>I think here it is about image size not number of Images. Percentages of public and private is not equal.</p>",
      "rawMarkdown": "I think here it is about image size not number of Images. Percentages of public and private is not equal.",
      "votes": null
    },
    {
      "id": "1327934",
      "postDate": "05/29/2021 18:42:03",
      "content": "<blockquote>\n  <p>All Images are in the test folder, otherwise how could you submit solution?</p>\n</blockquote>\n<p>I think the answer must be calculated inside Kaggle's notebook.<br>\nIf all test images are given, what's the puprose of notebooks? We could just send *.csv file.</p>",
      "rawMarkdown": "> All Images are in the test folder, otherwise how could you submit solution?\n\nI think the answer must be calculated inside Kaggle's notebook.\nIf all test images are given, what's the puprose of notebooks? We could just send *.csv file.",
      "votes": null
    },
    {
      "id": "1327936",
      "postDate": "05/29/2021 18:46:30",
      "content": "<p>This is a valid point, we can calculate output of our models for all images.</p>",
      "rawMarkdown": "This is a valid point, we can calculate output of our models for all images.",
      "votes": null
    },
    {
      "id": "1329692",
      "postDate": "05/31/2021 09:58:06",
      "content": "<p>I think I was wrong, because I assume the real test data is really \"hidden\".<br>\nBut in that case it is important to note that all the preprocessing needs to be done in the kernel, so all the PNG or JPG datasets with test are useless.</p>",
      "rawMarkdown": "I think I was wrong, because I assume the real test data is really \"hidden\".\nBut in that case it is important to note that all the preprocessing needs to be done in the kernel, so all the PNG or JPG datasets with test are useless.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1316178,
      "author_name": "osciiart",
      "author_url": "",
      "post_date": "05/20/2021 11:02:06",
      "content": "<p>It's mentioned in <a href=\"https://www.kaggle.com/c/siim-covid19-detection/data\" target=\"_blank\">Data</a>.<br>\n<code>The hidden test dataset is of roughly the same scale as the training dataset.</code></p>",
      "votes": null,
      "replies": [
        {
          "id": 1327808,
          "author_name": "jacekpoplawski",
          "author_url": "",
          "post_date": "05/29/2021 16:28:15",
          "content": "<p>I think here it is about image size not number of Images. Percentages of public and private is not equal.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1327710,
      "author_name": "sergeyzlobin",
      "author_url": "",
      "post_date": "05/29/2021 14:57:25",
      "content": "<p>The images that are in the \"test\" folder. Are they from the public part of the test dataset?</p>",
      "votes": null,
      "replies": [
        {
          "id": 1327803,
          "author_name": "jacekpoplawski",
          "author_url": "",
          "post_date": "05/29/2021 16:26:22",
          "content": "<p>All Images are in the test folder, otherwise how could you submit solution? Images are in dicom format but most people convert them into jpg or png so If there are any more Images than in test folder they won’t be converted and couldn’t be used in the kernel.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1327934,
          "author_name": "sergeyzlobin",
          "author_url": "",
          "post_date": "05/29/2021 18:42:03",
          "content": "<blockquote>\n  <p>All Images are in the test folder, otherwise how could you submit solution?</p>\n</blockquote>\n<p>I think the answer must be calculated inside Kaggle's notebook.<br>\nIf all test images are given, what's the puprose of notebooks? We could just send *.csv file.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1327936,
          "author_name": "jacekpoplawski",
          "author_url": "",
          "post_date": "05/29/2021 18:46:30",
          "content": "<p>This is a valid point, we can calculate output of our models for all images.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1329692,
          "author_name": "jacekpoplawski",
          "author_url": "",
          "post_date": "05/31/2021 09:58:06",
          "content": "<p>I think I was wrong, because I assume the real test data is really \"hidden\".<br>\nBut in that case it is important to note that all the preprocessing needs to be done in the kernel, so all the PNG or JPG datasets with test are useless.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1315841": "I don't see it mentioned anywhere; or am I just missing something obvious?\nThanks",
    "1316178": "It's mentioned in [Data](https://www.kaggle.com/c/siim-covid19-detection/data).\n`The hidden test dataset is of roughly the same scale as the training dataset.`",
    "1327710": "The images that are in the \"test\" folder. Are they from the public part of the test dataset?",
    "1327803": "All Images are in the test folder, otherwise how could you submit solution? Images are in dicom format but most people convert them into jpg or png so If there are any more Images than in test folder they won’t be converted and couldn’t be used in the kernel.",
    "1327808": "I think here it is about image size not number of Images. Percentages of public and private is not equal.",
    "1327934": "> All Images are in the test folder, otherwise how could you submit solution?\n\nI think the answer must be calculated inside Kaggle's notebook.\nIf all test images are given, what's the puprose of notebooks? We could just send *.csv file.",
    "1327936": "This is a valid point, we can calculate output of our models for all images.",
    "1329692": "I think I was wrong, because I assume the real test data is really \"hidden\".\nBut in that case it is important to note that all the preprocessing needs to be done in the kernel, so all the PNG or JPG datasets with test are useless."
  },
  "source": "meta"
}