{
  "id": 212613,
  "title": "How does the Scoring and Submission Process work?",
  "url": "/competitions/hubmap-kidney-segmentation/discussion/212613",
  "author_name": "Mark A Lavin",
  "post_date": "2021-01-19T14:40:35.992000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I am confused about how the Scoring and Submission process work.   There are three data sets:   Train, Public Test and Private Test.   My understanding is that for an \"official\" submission, one that is eligible for the competition, we submit our <strong>program</strong>, and the Submission mechanism runs the program on the Private Test dataset, not revealing that dataset nor the details of the run, only a final status (e.g., Timed Out) and a score.</p>\n<p>So…  I'm wondering, what do we do with the Public Test dataset?   Is there some sort of \"pre-Submission\" where I would submit the result of running my program on the Public Test data set, not the program itself, and the Submission mechanism scores those results against ground truth masks  that it holds but does not disclose for the Public Test dataset.</p>\n<p>Also, a question about the \"final Submission\":    My solution includes preprocessing of the input images to break them into tiles.   Usually, I do this separately from the analysis of the preprocessed images.   I gather that the program I submit must do the preprocessing of the Private Test dataset along with the prediction and encoding, storing the preprocessing results in <code>kaggle/working</code>.   Is that correct?</p>",
  "messages": [
    {
      "id": 1159915,
      "postDate": "2021-01-19T14:40:35.993Z",
      "content": "<p>I am confused about how the Scoring and Submission process work.   There are three data sets:   Train, Public Test and Private Test.   My understanding is that for an \"official\" submission, one that is eligible for the competition, we submit our <strong>program</strong>, and the Submission mechanism runs the program on the Private Test dataset, not revealing that dataset nor the details of the run, only a final status (e.g., Timed Out) and a score.</p>\n<p>So…  I'm wondering, what do we do with the Public Test dataset?   Is there some sort of \"pre-Submission\" where I would submit the result of running my program on the Public Test data set, not the program itself, and the Submission mechanism scores those results against ground truth masks  that it holds but does not disclose for the Public Test dataset.</p>\n<p>Also, a question about the \"final Submission\":    My solution includes preprocessing of the input images to break them into tiles.   Usually, I do this separately from the analysis of the preprocessed images.   I gather that the program I submit must do the preprocessing of the Private Test dataset along with the prediction and encoding, storing the preprocessing results in <code>kaggle/working</code>.   Is that correct?</p>",
      "rawMarkdown": "I am confused about how the Scoring and Submission process work.   There are three data sets:   Train, Public Test and Private Test.   My understanding is that for an \"official\" submission, one that is eligible for the competition, we submit our **program**, and the Submission mechanism runs the program on the Private Test dataset, not revealing that dataset nor the details of the run, only a final status (e.g., Timed Out) and a score.\n\nSo...  I'm wondering, what do we do with the Public Test dataset?   Is there some sort of \"pre-Submission\" where I would submit the result of running my program on the Public Test data set, not the program itself, and the Submission mechanism scores those results against ground truth masks  that it holds but does not disclose for the Public Test dataset.\n\nAlso, a question about the \"final Submission\":    My solution includes preprocessing of the input images to break them into tiles.   Usually, I do this separately from the analysis of the preprocessed images.   I gather that the program I submit must do the preprocessing of the Private Test dataset along with the prediction and encoding, storing the preprocessing results in ```kaggle/working```.   Is that correct?",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1159915": "I am confused about how the Scoring and Submission process work.   There are three data sets:   Train, Public Test and Private Test.   My understanding is that for an \"official\" submission, one that is eligible for the competition, we submit our **program**, and the Submission mechanism runs the program on the Private Test dataset, not revealing that dataset nor the details of the run, only a final status (e.g., Timed Out) and a score.\n\nSo...  I'm wondering, what do we do with the Public Test dataset?   Is there some sort of \"pre-Submission\" where I would submit the result of running my program on the Public Test data set, not the program itself, and the Submission mechanism scores those results against ground truth masks  that it holds but does not disclose for the Public Test dataset.\n\nAlso, a question about the \"final Submission\":    My solution includes preprocessing of the input images to break them into tiles.   Usually, I do this separately from the analysis of the preprocessed images.   I gather that the program I submit must do the preprocessing of the Private Test dataset along with the prediction and encoding, storing the preprocessing results in ```kaggle/working```.   Is that correct?"
  }
}