{
  "id": 104018,
  "title": "Are there corrupt images in the test set?",
  "url": "/competitions/aptos2019-blindness-detection/discussion/104018",
  "author_name": "",
  "post_date": "2019-08-13T16:37:04.682306400Z",
  "votes": 5,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Yesterday I was not able to submit to the competition because I was getting <code>Submission CSV not found</code> error. I posted it <a href=\"https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/103901#latest-598444\">here</a>. </p>\n\n<p>I submitted the kernel now with just one change and everything went smoothly. </p>\n\n<p>What did I change?\nI put a <code>try...except</code> block for the image reading part. </p>\n\n<p>This suggests that there may be some corrupt images in the test set. I have used all of my submissions for today, so can't validate it today. But If someone does, please report it here so that everyone is on the same page regarding this bug. </p>",
  "messages": [
    {
      "id": "598495",
      "postDate": "08/13/2019 16:37:04",
      "content": "<p>Yesterday I was not able to submit to the competition because I was getting <code>Submission CSV not found</code> error. I posted it <a href=\"https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/103901#latest-598444\">here</a>. </p>\n\n<p>I submitted the kernel now with just one change and everything went smoothly. </p>\n\n<p>What did I change?\nI put a <code>try...except</code> block for the image reading part. </p>\n\n<p>This suggests that there may be some corrupt images in the test set. I have used all of my submissions for today, so can't validate it today. But If someone does, please report it here so that everyone is on the same page regarding this bug. </p>",
      "rawMarkdown": "Yesterday I was not able to submit to the competition because I was getting `Submission CSV not found` error. I posted it [here](https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/103901#latest-598444). \n\nI submitted the kernel now with just one change and everything went smoothly. \n\nWhat did I change?\nI put a `try...except` block for the image reading part. \n\nThis suggests that there may be some corrupt images in the test set. I have used all of my submissions for today, so can't validate it today. But If someone does, please report it here so that everyone is on the same page regarding this bug.",
      "votes": null
    },
    {
      "id": "598624",
      "postDate": "08/13/2019 19:52:10",
      "content": "<p>If you are using one of the cropping methods from the kernels, if the picture is fully black, it will return an empty mask and cause errors. This may or may not be your error.</p>\n\n<p>I doubt any of the images are corrupted. Submission process for most of us have gone without error.</p>",
      "rawMarkdown": "If you are using one of the cropping methods from the kernels, if the picture is fully black, it will return an empty mask and cause errors. This may or may not be your error.\n\nI doubt any of the images are corrupted. Submission process for most of us have gone without error.",
      "votes": null
    },
    {
      "id": "598842",
      "postDate": "08/14/2019 05:48:03",
      "content": "<p>Hmm. Although I am not using the kernel based preprocessing but yeah, a full dark black image can be problematic in my case as well. Thanks </p>",
      "rawMarkdown": "Hmm. Although I am not using the kernel based preprocessing but yeah, a full dark black image can be problematic in my case as well. Thanks",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 598624,
      "author_name": "tanlikesmath",
      "author_url": "",
      "post_date": "08/13/2019 19:52:10",
      "content": "<p>If you are using one of the cropping methods from the kernels, if the picture is fully black, it will return an empty mask and cause errors. This may or may not be your error.</p>\n\n<p>I doubt any of the images are corrupted. Submission process for most of us have gone without error.</p>",
      "votes": null,
      "replies": [
        {
          "id": 598842,
          "author_name": "aakashnain",
          "author_url": "",
          "post_date": "08/14/2019 05:48:03",
          "content": "<p>Hmm. Although I am not using the kernel based preprocessing but yeah, a full dark black image can be problematic in my case as well. Thanks </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "598495": "Yesterday I was not able to submit to the competition because I was getting `Submission CSV not found` error. I posted it [here](https://www.kaggle.com/c/aptos2019-blindness-detection/discussion/103901#latest-598444). \n\nI submitted the kernel now with just one change and everything went smoothly. \n\nWhat did I change?\nI put a `try...except` block for the image reading part. \n\nThis suggests that there may be some corrupt images in the test set. I have used all of my submissions for today, so can't validate it today. But If someone does, please report it here so that everyone is on the same page regarding this bug.",
    "598624": "If you are using one of the cropping methods from the kernels, if the picture is fully black, it will return an empty mask and cause errors. This may or may not be your error.\n\nI doubt any of the images are corrupted. Submission process for most of us have gone without error.",
    "598842": "Hmm. Although I am not using the kernel based preprocessing but yeah, a full dark black image can be problematic in my case as well. Thanks"
  },
  "source": "meta"
}