{
  "id": 5880,
  "title": "HAND LABELING",
  "url": "/competitions/dogs-vs-cats/discussion/5880",
  "author_name": "",
  "post_date": "2013-09-25T22:41:34.550Z",
  "votes": null,
  "comment_count": 8,
  "views": 4994,
  "content": "<p>&quot;HAND LABELING IMAGES IS STRICTLY FORBIDDEN&quot; does it mean that we are not allowed to further refine training data by explicitly outlining regions on training images where object is present? </p>",
  "messages": [
    {
      "id": "31571",
      "postDate": "09/25/2013 22:41:34",
      "content": "<p>&quot;HAND LABELING IMAGES IS STRICTLY FORBIDDEN&quot; does it mean that we are not allowed to further refine training data by explicitly outlining regions on training images where object is present? </p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "31572",
      "postDate": "09/25/2013 22:46:05",
      "content": "<p>Yes, that would not be allowed. The algorithm should be capable of being trained and tested without human intervention.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "31592",
      "postDate": "09/26/2013 11:43:06",
      "content": "<p>Can we use more test data, though?</p>\n<p>&nbsp;</p>\n<p>As in: Test data from outside sources (cat pics, dog pictures)</p>\n<p>&nbsp;</p>\n<p>Marcus</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "31594",
      "postDate": "09/26/2013 12:12:51",
      "content": "<p>Yes, in the rules:</p>\n<p><code>External data is allowed, but must be posted in the forums for all to see and use.<br></code></p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "31598",
      "postDate": "09/26/2013 12:21:47",
      "content": "<p>&lt;&lt;&lt; im using that picture</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "31604",
      "postDate": "09/26/2013 12:49:19",
      "content": "<p><code>ncats = ncats + 1;</code></p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32470",
      "postDate": "10/20/2013 21:47:15",
      "content": "<p>Just browsing through the database looking at the first in my random selection, and I noticed this image, which is not of a cat but of some dude. &nbsp;He looks vaguely familiar but I can't remember who he is. &nbsp;I don't think he's Cat Stevens (Yusuf Islam).</p>\n<p>cat.7377.jpg</p>\n<p><span style=\"line-height: 1.4\">ncats -= 1</span></p>\n<p>:-)</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "36169",
      "postDate": "12/14/2013 11:54:30",
      "content": "<p>Thanks for the poster and answers. I thought it was a great idea to first crop the animal heads by hand and then train the model to detect cat/dog head in the image. Seems not possible.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "36234",
      "postDate": "12/17/2013 05:32:10",
      "content": "<p>Jie Liu, I thought the same</p>\n<p>Almost all the implementation for object detection first crop the image and highlight the object to be used and then train &nbsp;the model</p>\n<p>Now I am lost on what/how to do this</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 31572,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "09/25/2013 22:46:05",
      "content": "<p>Yes, that would not be allowed. The algorithm should be capable of being trained and tested without human intervention.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 31592,
      "author_name": "marcusklinge",
      "author_url": "",
      "post_date": "09/26/2013 11:43:06",
      "content": "<p>Can we use more test data, though?</p>\n<p>&nbsp;</p>\n<p>As in: Test data from outside sources (cat pics, dog pictures)</p>\n<p>&nbsp;</p>\n<p>Marcus</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 31594,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "09/26/2013 12:12:51",
      "content": "<p>Yes, in the rules:</p>\n<p><code>External data is allowed, but must be posted in the forums for all to see and use.<br></code></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 31598,
      "author_name": "foobarbazquux",
      "author_url": "",
      "post_date": "09/26/2013 12:21:47",
      "content": "<p>&lt;&lt;&lt; im using that picture</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 31604,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "09/26/2013 12:49:19",
      "content": "<p><code>ncats = ncats + 1;</code></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32470,
      "author_name": "ericsaurus",
      "author_url": "",
      "post_date": "10/20/2013 21:47:15",
      "content": "<p>Just browsing through the database looking at the first in my random selection, and I noticed this image, which is not of a cat but of some dude. &nbsp;He looks vaguely familiar but I can't remember who he is. &nbsp;I don't think he's Cat Stevens (Yusuf Islam).</p>\n<p>cat.7377.jpg</p>\n<p><span style=\"line-height: 1.4\">ncats -= 1</span></p>\n<p>:-)</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 36169,
      "author_name": "jieliu1",
      "author_url": "",
      "post_date": "12/14/2013 11:54:30",
      "content": "<p>Thanks for the poster and answers. I thought it was a great idea to first crop the animal heads by hand and then train the model to detect cat/dog head in the image. Seems not possible.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 36234,
      "author_name": "nvedia",
      "author_url": "",
      "post_date": "12/17/2013 05:32:10",
      "content": "<p>Jie Liu, I thought the same</p>\n<p>Almost all the implementation for object detection first crop the image and highlight the object to be used and then train &nbsp;the model</p>\n<p>Now I am lost on what/how to do this</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "31571": "",
    "31572": "",
    "31592": "",
    "31594": "",
    "31598": "",
    "31604": "",
    "32470": "",
    "36169": "",
    "36234": ""
  },
  "source": "meta"
}