{
  "id": 16402,
  "title": "Whale detector got bad performance",
  "url": "/competitions/noaa-right-whale-recognition/discussion/16402",
  "author_name": "Runsheng Song",
  "post_date": "2015-09-09T21:18:45.677000",
  "votes": 5,
  "comment_count": 0,
  "views": 894,
  "content": "<p>Hi all, I was reading the threads in the forum and practiced the whale detector by two ways. Unfortunately both of them work badly. Any suggestion on how many positive and negative label do we need to get a better detector? Here is my way.</p>\n\n<ol>\n<li><p>Using the method introduced by the tutorial, with 100 positive labels and 1000 negative labels. </p></li>\n<li><p>Convert the picture to binary color with certain threshold, this gives me whale pictures in black and white (in the attachment), I think this should be very easy to detect but still it doesn't work very well.</p></li>\n</ol>\n\n<p>So I am wondering how many labels should I create for this. Any suggestion appreciated </p>",
  "messages": [
    {
      "id": 91968,
      "postDate": "2015-09-09T21:18:45.677Z",
      "content": "<p>Hi all, I was reading the threads in the forum and practiced the whale detector by two ways. Unfortunately both of them work badly. Any suggestion on how many positive and negative label do we need to get a better detector? Here is my way.</p>\n\n<ol>\n<li><p>Using the method introduced by the tutorial, with 100 positive labels and 1000 negative labels. </p></li>\n<li><p>Convert the picture to binary color with certain threshold, this gives me whale pictures in black and white (in the attachment), I think this should be very easy to detect but still it doesn't work very well.</p></li>\n</ol>\n\n<p>So I am wondering how many labels should I create for this. Any suggestion appreciated </p>",
      "rawMarkdown": "Hi all, I was reading the threads in the forum and practiced the whale detector by two ways. Unfortunately both of them work badly. Any suggestion on how many positive and negative label do we need to get a better detector? Here is my way.\r\n\r\n1. Using the method introduced by the tutorial, with 100 positive labels and 1000 negative labels. \r\n\r\n2. Convert the picture to binary color with certain threshold, this gives me whale pictures in black and white (in the attachment), I think this should be very easy to detect but still it doesn't work very well.\r\n\r\nSo I am wondering how many labels should I create for this. Any suggestion appreciated ",
      "votes": 5
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "91968": "Hi all, I was reading the threads in the forum and practiced the whale detector by two ways. Unfortunately both of them work badly. Any suggestion on how many positive and negative label do we need to get a better detector? Here is my way.\r\n\r\n1. Using the method introduced by the tutorial, with 100 positive labels and 1000 negative labels. \r\n\r\n2. Convert the picture to binary color with certain threshold, this gives me whale pictures in black and white (in the attachment), I think this should be very easy to detect but still it doesn't work very well.\r\n\r\nSo I am wondering how many labels should I create for this. Any suggestion appreciated "
  }
}