{
  "id": 332245,
  "title": "Post processing after model output",
  "url": "/competitions/denoising-shabby-pages/discussion/332245",
  "author_name": "",
  "post_date": "2022-06-21T01:34:01.290347500Z",
  "votes": null,
  "comment_count": 5,
  "views": 0,
  "content": "<p>HI,</p>\n<p>Is there anyone try with additional post processing methods such as median filter, dilation or erosion after getting the results from deep learning model?</p>\n<p>I wonder if that is able to further improve the competition result.</p>",
  "messages": [
    {
      "id": "1827297",
      "postDate": "06/21/2022 01:34:01",
      "content": "<p>HI,</p>\n<p>Is there anyone try with additional post processing methods such as median filter, dilation or erosion after getting the results from deep learning model?</p>\n<p>I wonder if that is able to further improve the competition result.</p>",
      "rawMarkdown": "HI,\n\nIs there anyone try with additional post processing methods such as median filter, dilation or erosion after getting the results from deep learning model?\n\nI wonder if that is able to further improve the competition result.",
      "votes": null
    },
    {
      "id": "1833049",
      "postDate": "06/25/2022 16:33:52",
      "content": "<p>I think we should checkout GAN models and how they do post processing. Preprocessing makes sense but post processing might not be a very good idea</p>",
      "rawMarkdown": "I think we should checkout GAN models and how they do post processing. Preprocessing makes sense but post processing might not be a very good idea",
      "votes": null
    },
    {
      "id": "1833723",
      "postDate": "06/26/2022 09:22:22",
      "content": "<p>Okay, i'm just wonder if that really help since i haven't try that. </p>",
      "rawMarkdown": "Okay, i'm just wonder if that really help since i haven't try that.",
      "votes": null
    },
    {
      "id": "1850633",
      "postDate": "07/10/2022 15:04:02",
      "content": "<p>See this notebook. <a href=\"https://www.kaggle.com/code/deepdreamx/simple-cv2-denoising\" target=\"_blank\">https://www.kaggle.com/code/deepdreamx/simple-cv2-denoising</a></p>",
      "rawMarkdown": "See this notebook. https://www.kaggle.com/code/deepdreamx/simple-cv2-denoising",
      "votes": null
    },
    {
      "id": "1854265",
      "postDate": "07/13/2022 14:45:21",
      "content": "<p>Hello <a href=\"https://www.kaggle.com/weisinx7\" target=\"_blank\">@weisinx7</a> <br>\nI tried this post processing (<a href=\"https://www.kaggle.com/code/shengyou222/denoising-shabby-pages-in-ocr-pix2pix-pytorch?scriptVersionId=100083241&amp;cellId=30\" target=\"_blank\">Image sharpness</a>) after pix2pix output.<br>\nBut it leads worse than the origin.</p>",
      "rawMarkdown": "Hello @weisinx7 \nI tried this post processing ([Image sharpness] (https://www.kaggle.com/code/shengyou222/denoising-shabby-pages-in-ocr-pix2pix-pytorch?scriptVersionId=100083241&cellId=30)) after pix2pix output.\nBut it leads worse than the origin.",
      "votes": null
    },
    {
      "id": "1854792",
      "postDate": "07/14/2022 02:53:31",
      "content": "<p>i see, have you try some other post processing method? For example, median filter, blob analysis to remove small connected pixels and etc?</p>",
      "rawMarkdown": "i see, have you try some other post processing method? For example, median filter, blob analysis to remove small connected pixels and etc?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1833049,
      "author_name": "deepdreamx",
      "author_url": "",
      "post_date": "06/25/2022 16:33:52",
      "content": "<p>I think we should checkout GAN models and how they do post processing. Preprocessing makes sense but post processing might not be a very good idea</p>",
      "votes": null,
      "replies": [
        {
          "id": 1833723,
          "author_name": "weisinx7",
          "author_url": "",
          "post_date": "06/26/2022 09:22:22",
          "content": "<p>Okay, i'm just wonder if that really help since i haven't try that. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1850633,
      "author_name": "deepdreamx",
      "author_url": "",
      "post_date": "07/10/2022 15:04:02",
      "content": "<p>See this notebook. <a href=\"https://www.kaggle.com/code/deepdreamx/simple-cv2-denoising\" target=\"_blank\">https://www.kaggle.com/code/deepdreamx/simple-cv2-denoising</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1854265,
      "author_name": "shengyou222",
      "author_url": "",
      "post_date": "07/13/2022 14:45:21",
      "content": "<p>Hello <a href=\"https://www.kaggle.com/weisinx7\" target=\"_blank\">@weisinx7</a> <br>\nI tried this post processing (<a href=\"https://www.kaggle.com/code/shengyou222/denoising-shabby-pages-in-ocr-pix2pix-pytorch?scriptVersionId=100083241&amp;cellId=30\" target=\"_blank\">Image sharpness</a>) after pix2pix output.<br>\nBut it leads worse than the origin.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1854792,
          "author_name": "weisinx7",
          "author_url": "",
          "post_date": "07/14/2022 02:53:31",
          "content": "<p>i see, have you try some other post processing method? For example, median filter, blob analysis to remove small connected pixels and etc?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1827297": "HI,\n\nIs there anyone try with additional post processing methods such as median filter, dilation or erosion after getting the results from deep learning model?\n\nI wonder if that is able to further improve the competition result.",
    "1833049": "I think we should checkout GAN models and how they do post processing. Preprocessing makes sense but post processing might not be a very good idea",
    "1833723": "Okay, i'm just wonder if that really help since i haven't try that.",
    "1850633": "See this notebook. https://www.kaggle.com/code/deepdreamx/simple-cv2-denoising",
    "1854265": "Hello @weisinx7 \nI tried this post processing ([Image sharpness] (https://www.kaggle.com/code/shengyou222/denoising-shabby-pages-in-ocr-pix2pix-pytorch?scriptVersionId=100083241&cellId=30)) after pix2pix output.\nBut it leads worse than the origin.",
    "1854792": "i see, have you try some other post processing method? For example, median filter, blob analysis to remove small connected pixels and etc?"
  },
  "source": "meta"
}