{
  "id": 412032,
  "title": "Can traditional non-deep-learning based tools do the trick? (background-subtraction -> thresholding)",
  "url": "/competitions/google-research-identify-contrails-reduce-global-warming/discussion/412032",
  "author_name": "Aryan Garg",
  "post_date": "2023-05-22T01:47:06.183000",
  "votes": 2,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Has someone tried anything along these lines? <br>\nPlease do share your findings!</p>",
  "messages": [
    {
      "id": 2269415,
      "postDate": "2023-05-22T13:15:59.237Z",
      "content": "<p><a href=\"https://amt.copernicus.org/articles/3/1089/2010/amt-3-1089-2010.pdf\" target=\"_blank\">This paper</a> might be of interest. It focuses on contrail tracking, rather than just segmentation, but segmentation is essentially a submodule of it. There is no deep learning involved. Note that it uses a different satellite than this dataset, so the bands are slightly different and the spatial and temporal resolution is lower.</p>",
      "rawMarkdown": "[This paper](https://amt.copernicus.org/articles/3/1089/2010/amt-3-1089-2010.pdf) might be of interest. It focuses on contrail tracking, rather than just segmentation, but segmentation is essentially a submodule of it. There is no deep learning involved. Note that it uses a different satellite than this dataset, so the bands are slightly different and the spatial and temporal resolution is lower.",
      "votes": 3,
      "replies": [
        {
          "id": 2269526,
          "postDate": "2023-05-22T14:18:16.643Z",
          "content": "<p>That's awesome! I'll make sure to implement it and mention you in credits 😄</p>",
          "rawMarkdown": "That's awesome! I'll make sure to implement it and mention you in credits 😄"
        }
      ]
    },
    {
      "id": 2269525,
      "postDate": "2023-05-22T14:18:09.390Z",
      "content": "<p>I get a .4 - .5 local CV with non-spatial analysis, i.e. intensities and thresholds.</p>",
      "rawMarkdown": "I get a .4 - .5 local CV with non-spatial analysis, i.e. intensities and thresholds.",
      "votes": 1,
      "replies": [
        {
          "id": 2269530,
          "postDate": "2023-05-22T14:19:40.723Z",
          "content": "<p>Not sure if I understand… </p>",
          "rawMarkdown": "Not sure if I understand... ",
          "replies": [
            {
              "id": 2269547,
              "postDate": "2023-05-22T14:28:43.910Z",
              "content": "<p>Just looking at pixels as isolated entities. Essentially as you write. A cross validation Dice score of 0.4 - 0.5 depending a bit.</p>",
              "rawMarkdown": "Just looking at pixels as isolated entities. Essentially as you write. A cross validation Dice score of 0.4 - 0.5 depending a bit."
            },
            {
              "id": 2269668,
              "postDate": "2023-05-22T16:05:26.460Z",
              "content": "<p>Oh awesome! Thanks :)<br>\nGood to know that I'm not alone in this boat!</p>",
              "rawMarkdown": "Oh awesome! Thanks :)\nGood to know that I'm not alone in this boat!"
            }
          ]
        }
      ]
    },
    {
      "id": 2269035,
      "postDate": "2023-05-22T07:27:18.963Z",
      "content": "<p>I mean that's pretty obvious but was just curious if anyone had tried anything along the lines..</p>",
      "rawMarkdown": "I mean that's pretty obvious but was just curious if anyone had tried anything along the lines..",
      "votes": 1
    },
    {
      "id": 2268766,
      "postDate": "2023-05-22T01:47:06.183Z",
      "content": "<p>Has someone tried anything along these lines? <br>\nPlease do share your findings!</p>",
      "rawMarkdown": "Has someone tried anything along these lines? \nPlease do share your findings!",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 2269415,
      "author_name": "Aaron Sarna",
      "author_url": "",
      "post_date": "2023-05-22T13:15:59.237000",
      "content": "<p><a href=\"https://amt.copernicus.org/articles/3/1089/2010/amt-3-1089-2010.pdf\" target=\"_blank\">This paper</a> might be of interest. It focuses on contrail tracking, rather than just segmentation, but segmentation is essentially a submodule of it. There is no deep learning involved. Note that it uses a different satellite than this dataset, so the bands are slightly different and the spatial and temporal resolution is lower.</p>",
      "votes": 3,
      "replies": [
        {
          "id": 2269526,
          "author_name": "Aryan Garg",
          "author_url": "",
          "post_date": "2023-05-22T14:18:16.643000",
          "content": "<p>That's awesome! I'll make sure to implement it and mention you in credits 😄</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 2269525,
      "author_name": "Tord Malmgren",
      "author_url": "",
      "post_date": "2023-05-22T14:18:09.390000",
      "content": "<p>I get a .4 - .5 local CV with non-spatial analysis, i.e. intensities and thresholds.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 2269530,
          "author_name": "Aryan Garg",
          "author_url": "",
          "post_date": "2023-05-22T14:19:40.723000",
          "content": "<p>Not sure if I understand… </p>",
          "votes": 0,
          "replies": [
            {
              "id": 2269547,
              "author_name": "Tord Malmgren",
              "author_url": "",
              "post_date": "2023-05-22T14:28:43.910000",
              "content": "<p>Just looking at pixels as isolated entities. Essentially as you write. A cross validation Dice score of 0.4 - 0.5 depending a bit.</p>",
              "votes": 0,
              "replies": []
            },
            {
              "id": 2269668,
              "author_name": "Aryan Garg",
              "author_url": "",
              "post_date": "2023-05-22T16:05:26.460000",
              "content": "<p>Oh awesome! Thanks :)<br>\nGood to know that I'm not alone in this boat!</p>",
              "votes": 0,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 2269035,
      "author_name": "Aryan Garg",
      "author_url": "",
      "post_date": "2023-05-22T07:27:18.963000",
      "content": "<p>I mean that's pretty obvious but was just curious if anyone had tried anything along the lines..</p>",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2269415": "[This paper](https://amt.copernicus.org/articles/3/1089/2010/amt-3-1089-2010.pdf) might be of interest. It focuses on contrail tracking, rather than just segmentation, but segmentation is essentially a submodule of it. There is no deep learning involved. Note that it uses a different satellite than this dataset, so the bands are slightly different and the spatial and temporal resolution is lower.",
    "2269525": "I get a .4 - .5 local CV with non-spatial analysis, i.e. intensities and thresholds.",
    "2269035": "I mean that's pretty obvious but was just curious if anyone had tried anything along the lines..",
    "2268766": "Has someone tried anything along these lines? \nPlease do share your findings!"
  }
}