{
  "id": 108032,
  "title": "Circle to Rectangle Preprocessing",
  "url": "/competitions/aptos2019-blindness-detection/discussion/108032",
  "author_name": "",
  "post_date": "2019-09-08T15:44:29.120033900Z",
  "votes": 27,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi all I just wanted to share my preprocessing function. The main idea is to remove the black area without loosing information. To do it I developed a circle to rectangle morph function. </p>\n\n<p>To do the task its necessary to find the black borders row-by-row and stretch the non-black pixels to the image borders. This kind of normalization showed an improvement of 0.01+ in our models. </p>\n\n<p>Take a look at my <a href=\"https://www.kaggle.com/titericz/circle-to-rectagle-preprocessing-1?scriptVersionId=20319364\">kernel</a></p>",
  "messages": [
    {
      "id": "621506",
      "postDate": "09/08/2019 15:44:29",
      "content": "<p>Hi all I just wanted to share my preprocessing function. The main idea is to remove the black area without loosing information. To do it I developed a circle to rectangle morph function. </p>\n\n<p>To do the task its necessary to find the black borders row-by-row and stretch the non-black pixels to the image borders. This kind of normalization showed an improvement of 0.01+ in our models. </p>\n\n<p>Take a look at my <a href=\"https://www.kaggle.com/titericz/circle-to-rectagle-preprocessing-1?scriptVersionId=20319364\">kernel</a></p>",
      "rawMarkdown": "Hi all I just wanted to share my preprocessing function. The main idea is to remove the black area without loosing information. To do it I developed a circle to rectangle morph function. \n\nTo do the task its necessary to find the black borders row-by-row and stretch the non-black pixels to the image borders. This kind of normalization showed an improvement of 0.01+ in our models. \n\nTake a look at my [kernel](https://www.kaggle.com/titericz/circle-to-rectagle-preprocessing-1?scriptVersionId=20319364)",
      "votes": null
    },
    {
      "id": "621724",
      "postDate": "09/08/2019 21:17:42",
      "content": "<p>I guess you're probably too busy to write kernels like this more often... but I'd really like to read more of your kernels :)\nthanks <a href=\"/titericz\">@titericz</a></p>",
      "rawMarkdown": "I guess you're probably too busy to write kernels like this more often... but I'd really like to read more of your kernels :)\nthanks @titericz",
      "votes": null
    },
    {
      "id": "621933",
      "postDate": "09/09/2019 05:24:39",
      "content": "<p>That's impressive! Thank you for sharing!</p>",
      "rawMarkdown": "That's impressive! Thank you for sharing!",
      "votes": null
    },
    {
      "id": "622631",
      "postDate": "09/09/2019 23:00:58",
      "content": "<p>i see you choose to not using ben's color by default. did ben's color preprocessing helps in your case?</p>",
      "rawMarkdown": "i see you choose to not using ben's color by default. did ben's color preprocessing helps in your case?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 621724,
      "author_name": "jesucristo",
      "author_url": "",
      "post_date": "09/08/2019 21:17:42",
      "content": "<p>I guess you're probably too busy to write kernels like this more often... but I'd really like to read more of your kernels :)\nthanks <a href=\"/titericz\">@titericz</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 621933,
      "author_name": "haqishen",
      "author_url": "",
      "post_date": "09/09/2019 05:24:39",
      "content": "<p>That's impressive! Thank you for sharing!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 622631,
      "author_name": "moewie94",
      "author_url": "",
      "post_date": "09/09/2019 23:00:58",
      "content": "<p>i see you choose to not using ben's color by default. did ben's color preprocessing helps in your case?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "621506": "Hi all I just wanted to share my preprocessing function. The main idea is to remove the black area without loosing information. To do it I developed a circle to rectangle morph function. \n\nTo do the task its necessary to find the black borders row-by-row and stretch the non-black pixels to the image borders. This kind of normalization showed an improvement of 0.01+ in our models. \n\nTake a look at my [kernel](https://www.kaggle.com/titericz/circle-to-rectagle-preprocessing-1?scriptVersionId=20319364)",
    "621724": "I guess you're probably too busy to write kernels like this more often... but I'd really like to read more of your kernels :)\nthanks @titericz",
    "621933": "That's impressive! Thank you for sharing!",
    "622631": "i see you choose to not using ben's color by default. did ben's color preprocessing helps in your case?"
  },
  "source": "meta"
}