{
  "id": 263612,
  "title": "Submission Scoring Error and no reason given ",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/263612",
  "author_name": "",
  "post_date": "2021-08-09T20:15:10.725059600Z",
  "votes": 3,
  "comment_count": 7,
  "views": 0,
  "content": "<p>I have tried a few variations now to get the submission.csv to be accepted by its still not working. </p>\n<p>My notebook runs and creates a submission.csv but after the submission eventually get back a scoring error with no further information. I have taken the sample csv as the starting point but no luck. </p>\n<p>I have compared that both the sample csv and my submission have the same column names and id values. I have even padded the ids with leading zeros as that is what the sample has. My submission now looks like:</p>\n<p>BraTS21ID, MGMT_value<br>\n00001, 0.5<br>\n00013, 0.5<br>\n00015, 0.5</p>\n<p>Except that the 0.5s are the actual predictions. I have tried both leaving my predictions (e.g. 0.68, 0.51, 0.35) as output and rounding them towards 1 or 0 (e.g. 1.0, 0.0, 1.0). </p>\n<p>I am not sure what else i can try since there is no information in the error. Any ideas from experienced Kagglers?</p>\n<p>The only thing I can think is that as I derive the labels from the folders of the test data perhaps there is something different about the structure of the LB test data?</p>",
  "messages": [
    {
      "id": "1462391",
      "postDate": "08/09/2021 20:15:10",
      "content": "<p>I have tried a few variations now to get the submission.csv to be accepted by its still not working. </p>\n<p>My notebook runs and creates a submission.csv but after the submission eventually get back a scoring error with no further information. I have taken the sample csv as the starting point but no luck. </p>\n<p>I have compared that both the sample csv and my submission have the same column names and id values. I have even padded the ids with leading zeros as that is what the sample has. My submission now looks like:</p>\n<p>BraTS21ID, MGMT_value<br>\n00001, 0.5<br>\n00013, 0.5<br>\n00015, 0.5</p>\n<p>Except that the 0.5s are the actual predictions. I have tried both leaving my predictions (e.g. 0.68, 0.51, 0.35) as output and rounding them towards 1 or 0 (e.g. 1.0, 0.0, 1.0). </p>\n<p>I am not sure what else i can try since there is no information in the error. Any ideas from experienced Kagglers?</p>\n<p>The only thing I can think is that as I derive the labels from the folders of the test data perhaps there is something different about the structure of the LB test data?</p>",
      "rawMarkdown": "I have tried a few variations now to get the submission.csv to be accepted by its still not working. \n\nMy notebook runs and creates a submission.csv but after the submission eventually get back a scoring error with no further information. I have taken the sample csv as the starting point but no luck. \n\nI have compared that both the sample csv and my submission have the same column names and id values. I have even padded the ids with leading zeros as that is what the sample has. My submission now looks like:\n\nBraTS21ID, MGMT_value\n00001, 0.5\n00013, 0.5\n00015, 0.5\n\nExcept that the 0.5s are the actual predictions. I have tried both leaving my predictions (e.g. 0.68, 0.51, 0.35) as output and rounding them towards 1 or 0 (e.g. 1.0, 0.0, 1.0). \n\nI am not sure what else i can try since there is no information in the error. Any ideas from experienced Kagglers?\n\nThe only thing I can think is that as I derive the labels from the folders of the test data perhaps there is something different about the structure of the LB test data?",
      "votes": null
    },
    {
      "id": "1462468",
      "postDate": "08/09/2021 21:14:50",
      "content": "<p>Your csv file must contain all of the hidden test set entries, even if only the public ones are used for now</p>",
      "rawMarkdown": "Your csv file must contain all of the hidden test set entries, even if only the public ones are used for now",
      "votes": null
    },
    {
      "id": "1462539",
      "postDate": "08/09/2021 22:18:45",
      "content": "<p>Is this a plain notebook just use a constant for predictions or with a model?<br>\nIf you used a model, please decode the dicom file first. I used the dicom to png dataset before, and when I try to use the dataset's test png files to submit, Kaggle gave me an error message as well. You need to use the test files from the competition dataset and convert them.</p>",
      "rawMarkdown": "Is this a plain notebook just use a constant for predictions or with a model?\nIf you used a model, please decode the dicom file first. I used the dicom to png dataset before, and when I try to use the dataset's test png files to submit, Kaggle gave me an error message as well. You need to use the test files from the competition dataset and convert them.",
      "votes": null
    },
    {
      "id": "1463346",
      "postDate": "08/10/2021 07:04:01",
      "content": "<p>Oh of course, I am using the png dataset and had forgotten that the original is in dicom. Thanks.</p>",
      "rawMarkdown": "Oh of course, I am using the png dataset and had forgotten that the original is in dicom. Thanks.",
      "votes": null
    },
    {
      "id": "1464675",
      "postDate": "08/10/2021 16:58:58",
      "content": "<p>Hey thanks for this response, helped me to get unblocked. I just realized I was also using the PNG dataset so will have to convert from DICOM!</p>",
      "rawMarkdown": "Hey thanks for this response, helped me to get unblocked. I just realized I was also using the PNG dataset so will have to convert from DICOM!",
      "votes": null
    },
    {
      "id": "1464926",
      "postDate": "08/10/2021 19:06:35",
      "content": "<p>A suggestion: make sure the converted image is as same as your png image. I use a dicom_to_png() function to convert dicom files. I compared the png NumPy array with the dicom NumPy array and make sure they are identical.  </p>",
      "rawMarkdown": "A suggestion: make sure the converted image is as same as your png image. I use a dicom_to_png() function to convert dicom files. I compared the png NumPy array with the dicom NumPy array and make sure they are identical.",
      "votes": null
    },
    {
      "id": "1467883",
      "postDate": "08/12/2021 07:05:18",
      "content": "<p>I have been trying to use the original dicom images as the input but fast-ai just seems to have too many bugs working with them. However I found the notebook failed with \"Notebook Exceeded Allowed Compute\" when submitting scores if I convert the images with this code <br>\n<code>for f in test_list:  \n    ds = pydicom.read_file(f)\n    img = ds.pixel_array\n    new_f = str(f).replace('../input/rsna-miccai-brain-tumor-radiogenomic-classification', 'rsna-miccai-png')\n    os.remove(new_f)\n    new_f = new_f.replace('.dcm', '.png')\n    cv2.imwrite(new_f, img)</code></p>\n<p>Is there a faster way to convert?</p>",
      "rawMarkdown": "I have been trying to use the original dicom images as the input but fast-ai just seems to have too many bugs working with them. However I found the notebook failed with \"Notebook Exceeded Allowed Compute\" when submitting scores if I convert the images with this code \n`for f in test_list:  \n    ds = pydicom.read_file(f)\n    img = ds.pixel_array\n    new_f = str(f).replace('../input/rsna-miccai-brain-tumor-radiogenomic-classification', 'rsna-miccai-png')\n    os.remove(new_f)\n    new_f = new_f.replace('.dcm', '.png')\n    cv2.imwrite(new_f, img)`\n\nIs there a faster way to convert?",
      "votes": null
    },
    {
      "id": "1537188",
      "postDate": "10/07/2021 09:39:36",
      "content": "<p>Hi There,<br>\nthis is my notebook link<br>\n<a href=\"https://www.kaggle.com/rahulbenal/rsnav10/notebook?scriptVersionId=76483620\" target=\"_blank\">https://www.kaggle.com/rahulbenal/rsnav10/notebook?scriptVersionId=76483620</a>  <br>\nit shows notebook exeption trew but in the log sats successfull run.<br>\nPlease help here.</p>",
      "rawMarkdown": "Hi There,\nthis is my notebook link\nhttps://www.kaggle.com/rahulbenal/rsnav10/notebook?scriptVersionId=76483620  \nit shows notebook exeption trew but in the log sats successfull run.\nPlease help here.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1462468,
      "author_name": "carloalbertobarbano",
      "author_url": "",
      "post_date": "08/09/2021 21:14:50",
      "content": "<p>Your csv file must contain all of the hidden test set entries, even if only the public ones are used for now</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1462539,
      "author_name": "bcghost",
      "author_url": "",
      "post_date": "08/09/2021 22:18:45",
      "content": "<p>Is this a plain notebook just use a constant for predictions or with a model?<br>\nIf you used a model, please decode the dicom file first. I used the dicom to png dataset before, and when I try to use the dataset's test png files to submit, Kaggle gave me an error message as well. You need to use the test files from the competition dataset and convert them.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1463346,
          "author_name": "moonshots",
          "author_url": "",
          "post_date": "08/10/2021 07:04:01",
          "content": "<p>Oh of course, I am using the png dataset and had forgotten that the original is in dicom. Thanks.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1464675,
          "author_name": "d223chen",
          "author_url": "",
          "post_date": "08/10/2021 16:58:58",
          "content": "<p>Hey thanks for this response, helped me to get unblocked. I just realized I was also using the PNG dataset so will have to convert from DICOM!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1464926,
          "author_name": "bcghost",
          "author_url": "",
          "post_date": "08/10/2021 19:06:35",
          "content": "<p>A suggestion: make sure the converted image is as same as your png image. I use a dicom_to_png() function to convert dicom files. I compared the png NumPy array with the dicom NumPy array and make sure they are identical.  </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1467883,
          "author_name": "moonshots",
          "author_url": "",
          "post_date": "08/12/2021 07:05:18",
          "content": "<p>I have been trying to use the original dicom images as the input but fast-ai just seems to have too many bugs working with them. However I found the notebook failed with \"Notebook Exceeded Allowed Compute\" when submitting scores if I convert the images with this code <br>\n<code>for f in test_list:  \n    ds = pydicom.read_file(f)\n    img = ds.pixel_array\n    new_f = str(f).replace('../input/rsna-miccai-brain-tumor-radiogenomic-classification', 'rsna-miccai-png')\n    os.remove(new_f)\n    new_f = new_f.replace('.dcm', '.png')\n    cv2.imwrite(new_f, img)</code></p>\n<p>Is there a faster way to convert?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1537188,
      "author_name": "rahulbenal",
      "author_url": "",
      "post_date": "10/07/2021 09:39:36",
      "content": "<p>Hi There,<br>\nthis is my notebook link<br>\n<a href=\"https://www.kaggle.com/rahulbenal/rsnav10/notebook?scriptVersionId=76483620\" target=\"_blank\">https://www.kaggle.com/rahulbenal/rsnav10/notebook?scriptVersionId=76483620</a>  <br>\nit shows notebook exeption trew but in the log sats successfull run.<br>\nPlease help here.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1462391": "I have tried a few variations now to get the submission.csv to be accepted by its still not working. \n\nMy notebook runs and creates a submission.csv but after the submission eventually get back a scoring error with no further information. I have taken the sample csv as the starting point but no luck. \n\nI have compared that both the sample csv and my submission have the same column names and id values. I have even padded the ids with leading zeros as that is what the sample has. My submission now looks like:\n\nBraTS21ID, MGMT_value\n00001, 0.5\n00013, 0.5\n00015, 0.5\n\nExcept that the 0.5s are the actual predictions. I have tried both leaving my predictions (e.g. 0.68, 0.51, 0.35) as output and rounding them towards 1 or 0 (e.g. 1.0, 0.0, 1.0). \n\nI am not sure what else i can try since there is no information in the error. Any ideas from experienced Kagglers?\n\nThe only thing I can think is that as I derive the labels from the folders of the test data perhaps there is something different about the structure of the LB test data?",
    "1462468": "Your csv file must contain all of the hidden test set entries, even if only the public ones are used for now",
    "1462539": "Is this a plain notebook just use a constant for predictions or with a model?\nIf you used a model, please decode the dicom file first. I used the dicom to png dataset before, and when I try to use the dataset's test png files to submit, Kaggle gave me an error message as well. You need to use the test files from the competition dataset and convert them.",
    "1463346": "Oh of course, I am using the png dataset and had forgotten that the original is in dicom. Thanks.",
    "1464675": "Hey thanks for this response, helped me to get unblocked. I just realized I was also using the PNG dataset so will have to convert from DICOM!",
    "1464926": "A suggestion: make sure the converted image is as same as your png image. I use a dicom_to_png() function to convert dicom files. I compared the png NumPy array with the dicom NumPy array and make sure they are identical.",
    "1467883": "I have been trying to use the original dicom images as the input but fast-ai just seems to have too many bugs working with them. However I found the notebook failed with \"Notebook Exceeded Allowed Compute\" when submitting scores if I convert the images with this code \n`for f in test_list:  \n    ds = pydicom.read_file(f)\n    img = ds.pixel_array\n    new_f = str(f).replace('../input/rsna-miccai-brain-tumor-radiogenomic-classification', 'rsna-miccai-png')\n    os.remove(new_f)\n    new_f = new_f.replace('.dcm', '.png')\n    cv2.imwrite(new_f, img)`\n\nIs there a faster way to convert?",
    "1537188": "Hi There,\nthis is my notebook link\nhttps://www.kaggle.com/rahulbenal/rsnav10/notebook?scriptVersionId=76483620  \nit shows notebook exeption trew but in the log sats successfull run.\nPlease help here."
  },
  "source": "meta"
}