{
  "id": 155054,
  "title": "Understanding Strides for OpenSlide Extraction",
  "url": "/competitions/prostate-cancer-grade-assessment/discussion/155054",
  "author_name": "Jaideep",
  "post_date": "2020-05-31T05:24:56.925000",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>In one of the public kernel using open slide for extracting patches i see  below peace of code is used to get subsequent x,y coordinates to start extraction of patches from.\n ```\nfor h in hcrops:</p>\n\n<pre><code>   for v in vcrops:\n\n      x_location = h*crop_size*down_samples[crop_level] # 1,4,8\n      y_location = v*crop_size*down_samples[crop_level]\n</code></pre>\n\n<p>```</p>\n\n<p>I tried altering downsamples[crop level ] to say just  2 or  with \n <code>x_location = h*crop_size+cropsize * 2</code></p>\n\n<p>but i get too many empty tiles for lot of images . can any one help how we can change downsamples to get better results. </p>",
  "messages": [
    {
      "id": 868261,
      "postDate": "2020-05-31T05:24:56.927Z",
      "content": "<p>In one of the public kernel using open slide for extracting patches i see  below peace of code is used to get subsequent x,y coordinates to start extraction of patches from.\n ```\nfor h in hcrops:</p>\n\n<pre><code>   for v in vcrops:\n\n      x_location = h*crop_size*down_samples[crop_level] # 1,4,8\n      y_location = v*crop_size*down_samples[crop_level]\n</code></pre>\n\n<p>```</p>\n\n<p>I tried altering downsamples[crop level ] to say just  2 or  with \n <code>x_location = h*crop_size+cropsize * 2</code></p>\n\n<p>but i get too many empty tiles for lot of images . can any one help how we can change downsamples to get better results. </p>",
      "rawMarkdown": "In one of the public kernel using open slide for extracting patches i see  below peace of code is used to get subsequent x,y coordinates to start extraction of patches from.\n ```\nfor h in hcrops:\n\n       for v in vcrops:\n\n          x_location = h*crop_size*down_samples[crop_level] # 1,4,8\n          y_location = v*crop_size*down_samples[crop_level]\n```\n\nI tried altering downsamples[crop level ] to say just  2 or  with \n `x_location = h*crop_size+cropsize * 2`\n\nbut i get too many empty tiles for lot of images . can any one help how we can change downsamples to get better results. \n\n ",
      "votes": 2
    },
    {
      "id": 869693,
      "postDate": "2020-06-01T08:02:27.380Z",
      "content": "<p>i got it  as it takes x,y of 0 level image as input</p>",
      "rawMarkdown": "i got it  as it takes x,y of 0 level image as input"
    }
  ],
  "comments": [
    {
      "id": 869693,
      "author_name": "Jaideep",
      "author_url": "",
      "post_date": "2020-06-01T08:02:27.380000",
      "content": "<p>i got it  as it takes x,y of 0 level image as input</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "868261": "In one of the public kernel using open slide for extracting patches i see  below peace of code is used to get subsequent x,y coordinates to start extraction of patches from.\n ```\nfor h in hcrops:\n\n       for v in vcrops:\n\n          x_location = h*crop_size*down_samples[crop_level] # 1,4,8\n          y_location = v*crop_size*down_samples[crop_level]\n```\n\nI tried altering downsamples[crop level ] to say just  2 or  with \n `x_location = h*crop_size+cropsize * 2`\n\nbut i get too many empty tiles for lot of images . can any one help how we can change downsamples to get better results. \n\n ",
    "869693": "i got it  as it takes x,y of 0 level image as input"
  }
}