{
  "id": 135125,
  "title": "The number of test data problem",
  "url": "/competitions/iwildcam-2020-fgvc7/discussion/135125",
  "author_name": "",
  "post_date": "2020-03-12T06:10:45.370416400Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Following Data Overview,</p>\n\n<p>&gt; The WCS training set contains 217,959 images from 441 locations, and the WCS test set contains 62,894 images from 111 locations. These 552 locations are spread across the globe.</p>\n\n<p>Test set contains 62,894 images. However, when I submit with this number of results I've got the evaluation exception.</p>\n\n<p>&gt; Evaluation Exception: Submission must have 62870 rows</p>\n\n<p>Sample submission file has '62870' rows.</p>\n\n<p>Maybe I should fit the number as sample submission. :(</p>\n\n<p>I spent almost all GPU quota already. I should wait one week more :(</p>",
  "messages": [
    {
      "id": "769666",
      "postDate": "03/12/2020 06:10:45",
      "content": "<p>Following Data Overview,</p>\n\n<p>&gt; The WCS training set contains 217,959 images from 441 locations, and the WCS test set contains 62,894 images from 111 locations. These 552 locations are spread across the globe.</p>\n\n<p>Test set contains 62,894 images. However, when I submit with this number of results I've got the evaluation exception.</p>\n\n<p>&gt; Evaluation Exception: Submission must have 62870 rows</p>\n\n<p>Sample submission file has '62870' rows.</p>\n\n<p>Maybe I should fit the number as sample submission. :(</p>\n\n<p>I spent almost all GPU quota already. I should wait one week more :(</p>",
      "rawMarkdown": "Following Data Overview,\n\n&gt; The WCS training set contains 217,959 images from 441 locations, and the WCS test set contains 62,894 images from 111 locations. These 552 locations are spread across the globe.\n\nTest set contains 62,894 images. However, when I submit with this number of results I've got the evaluation exception.\n\n&gt; Evaluation Exception: Submission must have 62870 rows\n\nSample submission file has '62870' rows.\n\nMaybe I should fit the number as sample submission. :(\n\nI spent almost all GPU quota already. I should wait one week more :(",
      "votes": null
    },
    {
      "id": "770085",
      "postDate": "03/12/2020 14:50:39",
      "content": "<p>Chanran, thanks for pointing this out! We've looked into it, and it appears that the libraries we were using just skipped corrupted files without throwing an error, so we missed that there were 24 corrupted images in the test set.  </p>\n\n<p>I'll make a discussion post about this and work with kaggle to remove the corrupted files. In the meantime, you can just remove the rows (by id) that don't appear in the sample submission!</p>",
      "rawMarkdown": "Chanran, thanks for pointing this out! We've looked into it, and it appears that the libraries we were using just skipped corrupted files without throwing an error, so we missed that there were 24 corrupted images in the test set.  \n\nI'll make a discussion post about this and work with kaggle to remove the corrupted files. In the meantime, you can just remove the rows (by id) that don't appear in the sample submission!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 770085,
      "author_name": "sbeery",
      "author_url": "",
      "post_date": "03/12/2020 14:50:39",
      "content": "<p>Chanran, thanks for pointing this out! We've looked into it, and it appears that the libraries we were using just skipped corrupted files without throwing an error, so we missed that there were 24 corrupted images in the test set.  </p>\n\n<p>I'll make a discussion post about this and work with kaggle to remove the corrupted files. In the meantime, you can just remove the rows (by id) that don't appear in the sample submission!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "769666": "Following Data Overview,\n\n&gt; The WCS training set contains 217,959 images from 441 locations, and the WCS test set contains 62,894 images from 111 locations. These 552 locations are spread across the globe.\n\nTest set contains 62,894 images. However, when I submit with this number of results I've got the evaluation exception.\n\n&gt; Evaluation Exception: Submission must have 62870 rows\n\nSample submission file has '62870' rows.\n\nMaybe I should fit the number as sample submission. :(\n\nI spent almost all GPU quota already. I should wait one week more :(",
    "770085": "Chanran, thanks for pointing this out! We've looked into it, and it appears that the libraries we were using just skipped corrupted files without throwing an error, so we missed that there were 24 corrupted images in the test set.  \n\nI'll make a discussion post about this and work with kaggle to remove the corrupted files. In the meantime, you can just remove the rows (by id) that don't appear in the sample submission!"
  },
  "source": "meta"
}