{
  "id": 64252,
  "title": "[FIXED] Evaluation Exception: Index was outside the bounds of the array.",
  "url": "/competitions/airbus-ship-detection/discussion/64252",
  "author_name": "",
  "post_date": "2018-08-27T09:40:09.831556600Z",
  "votes": 5,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Can anyone tell us what this error means? We had no problems with previous submissions using same pipeline.</p>\n\n<p>EDIT: We have also made sure to remove any pixels at the image border from predictions, and there are no overlapping ships.</p>",
  "messages": [
    {
      "id": "376350",
      "postDate": "08/27/2018 09:40:09",
      "content": "<p>Can anyone tell us what this error means? We had no problems with previous submissions using same pipeline.</p>\n\n<p>EDIT: We have also made sure to remove any pixels at the image border from predictions, and there are no overlapping ships.</p>",
      "rawMarkdown": "Can anyone tell us what this error means? We had no problems with previous submissions using same pipeline.\n\nEDIT: We have also made sure to remove any pixels at the image border from predictions, and there are no overlapping ships.",
      "votes": null
    },
    {
      "id": "376839",
      "postDate": "08/28/2018 06:36:16",
      "content": "<p>Could it be that you have  rle encoded one or more mask without any pixel rather than passing an empty string in your submission file ?</p>",
      "rawMarkdown": "Could it be that you have  rle encoded one or more mask without any pixel rather than passing an empty string in your submission file ?",
      "votes": null
    },
    {
      "id": "379408",
      "postDate": "08/31/2018 10:17:38",
      "content": "<p>Did you check that you do not have NaNs ?</p>",
      "rawMarkdown": "Did you check that you do not have NaNs ?",
      "votes": null
    },
    {
      "id": "379428",
      "postDate": "08/31/2018 10:48:20",
      "content": "<p>Thanks for help, we managed to found the problem - the cause was appending a row with NaN encoded pixels to an image that actually had predictions in previous rows. This is the code we used to check submissions:</p>\n\n<pre><code>import pandas as pd  \ndf = pd.read_csv(\"kaggle.csv\")\ngr = df.groupby(\"ImageId\")[\"EncodedPixels\"].apply(lambda x: x.isnull().any() and len(x) &gt; 1)  \nprint(gr.value_counts())  # should all be false  \nprint(gr[gr])  # these images have predictions and nan rows  \n</code></pre>",
      "rawMarkdown": "Thanks for help, we managed to found the problem - the cause was appending a row with NaN encoded pixels to an image that actually had predictions in previous rows. This is the code we used to check submissions:\n\n    import pandas as pd  \n    df = pd.read_csv(\"kaggle.csv\")\n    gr = df.groupby(\"ImageId\")[\"EncodedPixels\"].apply(lambda x: x.isnull().any() and len(x) &gt; 1)  \n    print(gr.value_counts())  # should all be false  \n    print(gr[gr])  # these images have predictions and nan rows",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 376839,
      "author_name": "alexiswoozcom",
      "author_url": "",
      "post_date": "08/28/2018 06:36:16",
      "content": "<p>Could it be that you have  rle encoded one or more mask without any pixel rather than passing an empty string in your submission file ?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 379408,
      "author_name": "benteo",
      "author_url": "",
      "post_date": "08/31/2018 10:17:38",
      "content": "<p>Did you check that you do not have NaNs ?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 379428,
      "author_name": "kowaalczyk",
      "author_url": "",
      "post_date": "08/31/2018 10:48:20",
      "content": "<p>Thanks for help, we managed to found the problem - the cause was appending a row with NaN encoded pixels to an image that actually had predictions in previous rows. This is the code we used to check submissions:</p>\n\n<pre><code>import pandas as pd  \ndf = pd.read_csv(\"kaggle.csv\")\ngr = df.groupby(\"ImageId\")[\"EncodedPixels\"].apply(lambda x: x.isnull().any() and len(x) &gt; 1)  \nprint(gr.value_counts())  # should all be false  \nprint(gr[gr])  # these images have predictions and nan rows  \n</code></pre>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "376350": "Can anyone tell us what this error means? We had no problems with previous submissions using same pipeline.\n\nEDIT: We have also made sure to remove any pixels at the image border from predictions, and there are no overlapping ships.",
    "376839": "Could it be that you have  rle encoded one or more mask without any pixel rather than passing an empty string in your submission file ?",
    "379408": "Did you check that you do not have NaNs ?",
    "379428": "Thanks for help, we managed to found the problem - the cause was appending a row with NaN encoded pixels to an image that actually had predictions in previous rows. This is the code we used to check submissions:\n\n    import pandas as pd  \n    df = pd.read_csv(\"kaggle.csv\")\n    gr = df.groupby(\"ImageId\")[\"EncodedPixels\"].apply(lambda x: x.isnull().any() and len(x) &gt; 1)  \n    print(gr.value_counts())  # should all be false  \n    print(gr[gr])  # these images have predictions and nan rows"
  },
  "source": "meta"
}