{
  "id": 50098,
  "title": "Why more accurate mapping gets a lower score ?",
  "url": "/competitions/data-science-bowl-2018/discussion/50098",
  "author_name": "",
  "post_date": "2018-02-20T01:45:09.075779200Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I've been struggling with my mapping algorithms, tried different approaches and none was good enough, so I thought maybe I need to figure out what's wrong with my algorithms by breaking the scores down to individual images and look at them one by one and try to find why I was getting low scores, so I tried to submit one image at a time, and see if I can adjust my parameters to find the best value for that image, meanwhile I can figure out what's the range for the score for one image, so far the best score I got for one image is 0.013, that's a reasonable value, if you do this :  0.013 * 65 = 0.845 , which is quite good, but then sometime strange happened.</p>\n\n<p>For image : 0a849e0eb15faa8a6d7329c3dd66aabe9a294cccb52ed30a90c8ca99092ae732</p>\n\n<p>I tried different parameters and the best score I can ever get is : 0.004</p>\n\n<p>And the strange thing is : a better mapped image got a lower score of only : 0.001</p>\n\n<p>Which is almost nothing, as you can see from the images below, the lower scored image is actually a better mapped image, so I wonder if there is something wrong with the scoring algorithm ?</p>\n\n<p><img src=\"http://nmjava.com/Score.PNG\" alt=\"enter image description here\"></p>",
  "messages": [
    {
      "id": "285340",
      "postDate": "02/20/2018 01:45:09",
      "content": "<p>I've been struggling with my mapping algorithms, tried different approaches and none was good enough, so I thought maybe I need to figure out what's wrong with my algorithms by breaking the scores down to individual images and look at them one by one and try to find why I was getting low scores, so I tried to submit one image at a time, and see if I can adjust my parameters to find the best value for that image, meanwhile I can figure out what's the range for the score for one image, so far the best score I got for one image is 0.013, that's a reasonable value, if you do this :  0.013 * 65 = 0.845 , which is quite good, but then sometime strange happened.</p>\n\n<p>For image : 0a849e0eb15faa8a6d7329c3dd66aabe9a294cccb52ed30a90c8ca99092ae732</p>\n\n<p>I tried different parameters and the best score I can ever get is : 0.004</p>\n\n<p>And the strange thing is : a better mapped image got a lower score of only : 0.001</p>\n\n<p>Which is almost nothing, as you can see from the images below, the lower scored image is actually a better mapped image, so I wonder if there is something wrong with the scoring algorithm ?</p>\n\n<p><img src=\"http://nmjava.com/Score.PNG\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "I've been struggling with my mapping algorithms, tried different approaches and none was good enough, so I thought maybe I need to figure out what's wrong with my algorithms by breaking the scores down to individual images and look at them one by one and try to find why I was getting low scores, so I tried to submit one image at a time, and see if I can adjust my parameters to find the best value for that image, meanwhile I can figure out what's the range for the score for one image, so far the best score I got for one image is 0.013, that's a reasonable value, if you do this :  0.013 * 65 = 0.845 , which is quite good, but then sometime strange happened.\n\nFor image : 0a849e0eb15faa8a6d7329c3dd66aabe9a294cccb52ed30a90c8ca99092ae732\n\nI tried different parameters and the best score I can ever get is : 0.004\n\nAnd the strange thing is : a better mapped image got a lower score of only : 0.001\n\nWhich is almost nothing, as you can see from the images below, the lower scored image is actually a better mapped image, so I wonder if there is something wrong with the scoring algorithm ?\n\n ![enter image description here][1]\n\n\n  [1]: http://nmjava.com/Score.PNG",
      "votes": null
    },
    {
      "id": "285355",
      "postDate": "02/20/2018 03:02:21",
      "content": "<p>further break down the score into each instance mask. label which  are tp and fp.</p>\n\n<p>analyse the precious at each iou threshold</p>\n\n<p>in your figure, the second image has better detection but worst segmentation.\nthe third image has worst detection but maybe better segmentation </p>",
      "rawMarkdown": "further break down the score into each instance mask. label which  are tp and fp.\n\nanalyse the precious at each iou threshold\n\n\nin your figure, the second image has better detection but worst segmentation.\nthe third image has worst detection but maybe better segmentation",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 285355,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "02/20/2018 03:02:21",
      "content": "<p>further break down the score into each instance mask. label which  are tp and fp.</p>\n\n<p>analyse the precious at each iou threshold</p>\n\n<p>in your figure, the second image has better detection but worst segmentation.\nthe third image has worst detection but maybe better segmentation </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "285340": "I've been struggling with my mapping algorithms, tried different approaches and none was good enough, so I thought maybe I need to figure out what's wrong with my algorithms by breaking the scores down to individual images and look at them one by one and try to find why I was getting low scores, so I tried to submit one image at a time, and see if I can adjust my parameters to find the best value for that image, meanwhile I can figure out what's the range for the score for one image, so far the best score I got for one image is 0.013, that's a reasonable value, if you do this :  0.013 * 65 = 0.845 , which is quite good, but then sometime strange happened.\n\nFor image : 0a849e0eb15faa8a6d7329c3dd66aabe9a294cccb52ed30a90c8ca99092ae732\n\nI tried different parameters and the best score I can ever get is : 0.004\n\nAnd the strange thing is : a better mapped image got a lower score of only : 0.001\n\nWhich is almost nothing, as you can see from the images below, the lower scored image is actually a better mapped image, so I wonder if there is something wrong with the scoring algorithm ?\n\n ![enter image description here][1]\n\n\n  [1]: http://nmjava.com/Score.PNG",
    "285355": "further break down the score into each instance mask. label which  are tp and fp.\n\nanalyse the precious at each iou threshold\n\n\nin your figure, the second image has better detection but worst segmentation.\nthe third image has worst detection but maybe better segmentation"
  },
  "source": "meta"
}