{
  "id": 471613,
  "title": "Struggling with very strange submission scoring error. Anyone has any hint?",
  "url": "/competitions/blood-vessel-segmentation/discussion/471613",
  "author_name": "",
  "post_date": "2024-01-28T22:15:40.608816800Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi there! I'm debugging my <strong>Submission Scoring Error</strong> and have read many of your discussion.<br>\n(<em>Your notebook generated a submission file with incorrect format. Some examples causing this are: wrong number of rows or columns, empty values, an incorrect data type for a value, or invalid submission values from what is expected. See more debugging tips</em>)</p>\n<p>I'm now having one \"dummy\" submission looping through all test images and simply predicts all rle as \"1 0\", succeeded. The other \"target\" submission where I comment out the prediction function and replace will assigning rle as \"1 0\", failed. This is causing me so much confusion.</p>\n<p>For sanity check I run the same code on the training data \"kidney_1_dense\" and \"kidney_3_sparse\" to mimic test data with 2 samples, and got 2 identical \"submission.csv\" output files. As the screenshot shows, to prove they are indeed identical I used <code>df.merge()</code> and the row count is exactly the sum of number of images in the 2 samples. (<code>id</code>s are each unique in each csv, checked as well.)</p>\n<p>I feel like hitting a dead end now so if anyone has any hint please help, thanks! T^T</p>",
  "messages": [
    {
      "id": "2624696",
      "postDate": "01/28/2024 22:15:40",
      "content": "<p>Hi there! I'm debugging my <strong>Submission Scoring Error</strong> and have read many of your discussion.<br>\n(<em>Your notebook generated a submission file with incorrect format. Some examples causing this are: wrong number of rows or columns, empty values, an incorrect data type for a value, or invalid submission values from what is expected. See more debugging tips</em>)</p>\n<p>I'm now having one \"dummy\" submission looping through all test images and simply predicts all rle as \"1 0\", succeeded. The other \"target\" submission where I comment out the prediction function and replace will assigning rle as \"1 0\", failed. This is causing me so much confusion.</p>\n<p>For sanity check I run the same code on the training data \"kidney_1_dense\" and \"kidney_3_sparse\" to mimic test data with 2 samples, and got 2 identical \"submission.csv\" output files. As the screenshot shows, to prove they are indeed identical I used <code>df.merge()</code> and the row count is exactly the sum of number of images in the 2 samples. (<code>id</code>s are each unique in each csv, checked as well.)</p>\n<p>I feel like hitting a dead end now so if anyone has any hint please help, thanks! T^T</p>",
      "rawMarkdown": "Hi there! I'm debugging my **Submission Scoring Error** and have read many of your discussion.\n(*Your notebook generated a submission file with incorrect format. Some examples causing this are: wrong number of rows or columns, empty values, an incorrect data type for a value, or invalid submission values from what is expected. See more debugging tips*)\n\nI'm now having one \"dummy\" submission looping through all test images and simply predicts all rle as \"1 0\", succeeded. The other \"target\" submission where I comment out the prediction function and replace will assigning rle as \"1 0\", failed. This is causing me so much confusion.\n\nFor sanity check I run the same code on the training data \"kidney_1_dense\" and \"kidney_3_sparse\" to mimic test data with 2 samples, and got 2 identical \"submission.csv\" output files. As the screenshot shows, to prove they are indeed identical I used `df.merge()` and the row count is exactly the sum of number of images in the 2 samples. (`id`s are each unique in each csv, checked as well.)\n\nI feel like hitting a dead end now so if anyone has any hint please help, thanks! T^T",
      "votes": null
    },
    {
      "id": "2624700",
      "postDate": "01/28/2024 22:28:15",
      "content": "<p>Make a completely general pipeline. Just assume that there will be some files at /test folder with some keys. Read them and use them. Don't try assume they will start at 0 etc… Use the same keys that your code reads.</p>\n<p>If you are preprocessing the inputs to feed your model (paddings, resizing,etc…) be sure you submit the mask in the original sizes.</p>",
      "rawMarkdown": "Make a completely general pipeline. Just assume that there will be some files at /test folder with some keys. Read them and use them. Don't try assume they will start at 0 etc... Use the same keys that your code reads.\n\nIf you are preprocessing the inputs to feed your model (paddings, resizing,etc...) be sure you submit the mask in the original sizes.",
      "votes": null
    },
    {
      "id": "2624764",
      "postDate": "01/29/2024 00:56:45",
      "content": "<p>Hi Angel!<br>\nThis is indeed the case! After all the preprocessing I've done I ended up using index rather than the image filename as is. It's a shame my debugging didn't catch this error. Now I finally got a working submission so thank you very much!</p>",
      "rawMarkdown": "Hi Angel!\nThis is indeed the case! After all the preprocessing I've done I ended up using index rather than the image filename as is. It's a shame my debugging didn't catch this error. Now I finally got a working submission so thank you very much!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2624700,
      "author_name": "sacuscreed",
      "author_url": "",
      "post_date": "01/28/2024 22:28:15",
      "content": "<p>Make a completely general pipeline. Just assume that there will be some files at /test folder with some keys. Read them and use them. Don't try assume they will start at 0 etc… Use the same keys that your code reads.</p>\n<p>If you are preprocessing the inputs to feed your model (paddings, resizing,etc…) be sure you submit the mask in the original sizes.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2624764,
          "author_name": "asphodel19150",
          "author_url": "",
          "post_date": "01/29/2024 00:56:45",
          "content": "<p>Hi Angel!<br>\nThis is indeed the case! After all the preprocessing I've done I ended up using index rather than the image filename as is. It's a shame my debugging didn't catch this error. Now I finally got a working submission so thank you very much!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2624696": "Hi there! I'm debugging my **Submission Scoring Error** and have read many of your discussion.\n(*Your notebook generated a submission file with incorrect format. Some examples causing this are: wrong number of rows or columns, empty values, an incorrect data type for a value, or invalid submission values from what is expected. See more debugging tips*)\n\nI'm now having one \"dummy\" submission looping through all test images and simply predicts all rle as \"1 0\", succeeded. The other \"target\" submission where I comment out the prediction function and replace will assigning rle as \"1 0\", failed. This is causing me so much confusion.\n\nFor sanity check I run the same code on the training data \"kidney_1_dense\" and \"kidney_3_sparse\" to mimic test data with 2 samples, and got 2 identical \"submission.csv\" output files. As the screenshot shows, to prove they are indeed identical I used `df.merge()` and the row count is exactly the sum of number of images in the 2 samples. (`id`s are each unique in each csv, checked as well.)\n\nI feel like hitting a dead end now so if anyone has any hint please help, thanks! T^T",
    "2624700": "Make a completely general pipeline. Just assume that there will be some files at /test folder with some keys. Read them and use them. Don't try assume they will start at 0 etc... Use the same keys that your code reads.\n\nIf you are preprocessing the inputs to feed your model (paddings, resizing,etc...) be sure you submit the mask in the original sizes.",
    "2624764": "Hi Angel!\nThis is indeed the case! After all the preprocessing I've done I ended up using index rather than the image filename as is. It's a shame my debugging didn't catch this error. Now I finally got a working submission so thank you very much!"
  },
  "source": "meta"
}