{
  "id": 12301,
  "title": "Skewed images",
  "url": "/competitions/datasciencebowl/discussion/12301",
  "author_name": "",
  "post_date": "2015-01-22T03:42:55.073Z",
  "votes": null,
  "comment_count": 1,
  "views": 941,
  "content": "<p>Are all images skewed. All images look perfectly to be &nbsp;from a side angle or the top.</p>\n<p>If so that is great !!</p>\n<p>If not may be i did not see enough.&nbsp;</p>\n<p>thanks</p>",
  "messages": [
    {
      "id": "62995",
      "postDate": "01/22/2015 03:42:55",
      "content": "<p>Are all images skewed. All images look perfectly to be &nbsp;from a side angle or the top.</p>\n<p>If so that is great !!</p>\n<p>If not may be i did not see enough.&nbsp;</p>\n<p>thanks</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "63017",
      "postDate": "01/22/2015 16:43:43",
      "content": "<p>I suspect it's the nature of how the dataset is generated that&nbsp;causes most of the images to either be from the top or the side and generally at more convenient angles for identification purposes. &nbsp;Each of these images in the training and testing datasets had to be classified and it's likely they were classified using human eyes. &nbsp;Images of&nbsp;plankton and whatnot at angles that even taxonomy experts can't identify would have to be put in the &quot;miscellaneous/unclassified&quot; category. &nbsp;Those that remain that could be classified are thus the ones that are at a convenient angle where classification is humanly possible.</p>\n<p>I really don't know for sure but reading up the description regarding where this data is coming from (https://www.kaggle.com/c/datasciencebowl) this explanation sounds as good as any.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 63017,
      "author_name": "sjfcannon",
      "author_url": "",
      "post_date": "01/22/2015 16:43:43",
      "content": "<p>I suspect it's the nature of how the dataset is generated that&nbsp;causes most of the images to either be from the top or the side and generally at more convenient angles for identification purposes. &nbsp;Each of these images in the training and testing datasets had to be classified and it's likely they were classified using human eyes. &nbsp;Images of&nbsp;plankton and whatnot at angles that even taxonomy experts can't identify would have to be put in the &quot;miscellaneous/unclassified&quot; category. &nbsp;Those that remain that could be classified are thus the ones that are at a convenient angle where classification is humanly possible.</p>\n<p>I really don't know for sure but reading up the description regarding where this data is coming from (https://www.kaggle.com/c/datasciencebowl) this explanation sounds as good as any.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "62995": "",
    "63017": ""
  },
  "source": "meta"
}