{
  "id": 37178,
  "title": "Background subtraction libraries and algorithms",
  "url": "/competitions/carvana-image-masking-challenge/discussion/37178",
  "author_name": "",
  "post_date": "2017-07-28T10:35:47.271880Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi,</p>\n\n<p>\"Currently the library offers 37 BGS algorithms.\" :</p>\n\n<p><a href=\"https://github.com/myfavouritekk/bgslibrary\">https://github.com/myfavouritekk/bgslibrary</a>\n<a href=\"https://github.com/andrewssobral/bgslibrary\">https://github.com/andrewssobral/bgslibrary</a>\nTrying to figure out if it can work easily <strong>on still images.</strong> </p>\n\n<p>\"A comprehensive review of background subtraction algorithms\":</p>\n\n<p><a href=\"https://www.researchgate.net/publication/259340906_A_comprehensive_review_of_background_subtraction_algorithms_evaluated_with_synthetic_and_real_videos\">https://www.researchgate.net/publication/259340906_A_comprehensive_review_of_background_subtraction_algorithms_evaluated_with_synthetic_and_real_videos</a></p>\n\n<p>Would love to read more if you have something interesting. </p>\n\n<p>Best, </p>",
  "messages": [
    {
      "id": "208038",
      "postDate": "07/28/2017 10:35:47",
      "content": "<p>Hi,</p>\n\n<p>\"Currently the library offers 37 BGS algorithms.\" :</p>\n\n<p><a href=\"https://github.com/myfavouritekk/bgslibrary\">https://github.com/myfavouritekk/bgslibrary</a>\n<a href=\"https://github.com/andrewssobral/bgslibrary\">https://github.com/andrewssobral/bgslibrary</a>\nTrying to figure out if it can work easily <strong>on still images.</strong> </p>\n\n<p>\"A comprehensive review of background subtraction algorithms\":</p>\n\n<p><a href=\"https://www.researchgate.net/publication/259340906_A_comprehensive_review_of_background_subtraction_algorithms_evaluated_with_synthetic_and_real_videos\">https://www.researchgate.net/publication/259340906_A_comprehensive_review_of_background_subtraction_algorithms_evaluated_with_synthetic_and_real_videos</a></p>\n\n<p>Would love to read more if you have something interesting. </p>\n\n<p>Best, </p>",
      "rawMarkdown": "Hi,\n\n\"Currently the library offers 37 BGS algorithms.\" :\n\nhttps://github.com/myfavouritekk/bgslibrary\nhttps://github.com/andrewssobral/bgslibrary\nTrying to figure out if it can work easily **on still images.** \n\n\"A comprehensive review of background subtraction algorithms\":\n\nhttps://www.researchgate.net/publication/259340906_A_comprehensive_review_of_background_subtraction_algorithms_evaluated_with_synthetic_and_real_videos\n\nWould love to read more if you have something interesting. \n\nBest,",
      "votes": null
    },
    {
      "id": "208040",
      "postDate": "07/28/2017 10:45:41",
      "content": "<p>Each vehicle has 16 frames of rotation, so within each set, the background should be static (save for some minor JPG noise). This may lend to techniques that require more frames. </p>\n\n<p>Some algorithms depend on a reference frame, and unfortunately we can't offer a reference frame without a vehicle, as such a photo is not taken for each vehicle. Although rare, there can be slight adjustments made to the camera, white balance/exposure (not sure if this is automatically optimized per-vehicle), and such over time.</p>\n\n<p>One word of caution when approaching background subtraction, though, is that reflective objects can be problematic, as pixels on the object can end up being the same color as the background was when the object was not there.</p>\n\n<p>This isn't an example of this effect specifically, but it illustrates pixel color similarities, and the areas in the frame that ever experience changes while the vehicle rotates:</p>\n\n<p><img src=\"http://i.imgur.com/HjwOtxX.png\" alt=\"enter image description here\" title=\"\"></p>",
      "rawMarkdown": "Each vehicle has 16 frames of rotation, so within each set, the background should be static (save for some minor JPG noise). This may lend to techniques that require more frames. \n\nSome algorithms depend on a reference frame, and unfortunately we can't offer a reference frame without a vehicle, as such a photo is not taken for each vehicle. Although rare, there can be slight adjustments made to the camera, white balance/exposure (not sure if this is automatically optimized per-vehicle), and such over time.\n\nOne word of caution when approaching background subtraction, though, is that reflective objects can be problematic, as pixels on the object can end up being the same color as the background was when the object was not there.\n\nThis isn't an example of this effect specifically, but it illustrates pixel color similarities, and the areas in the frame that ever experience changes while the vehicle rotates:\n\n![enter image description here][1]\n\n  [1]: http://i.imgur.com/HjwOtxX.png",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 208040,
      "author_name": "brianshaler",
      "author_url": "",
      "post_date": "07/28/2017 10:45:41",
      "content": "<p>Each vehicle has 16 frames of rotation, so within each set, the background should be static (save for some minor JPG noise). This may lend to techniques that require more frames. </p>\n\n<p>Some algorithms depend on a reference frame, and unfortunately we can't offer a reference frame without a vehicle, as such a photo is not taken for each vehicle. Although rare, there can be slight adjustments made to the camera, white balance/exposure (not sure if this is automatically optimized per-vehicle), and such over time.</p>\n\n<p>One word of caution when approaching background subtraction, though, is that reflective objects can be problematic, as pixels on the object can end up being the same color as the background was when the object was not there.</p>\n\n<p>This isn't an example of this effect specifically, but it illustrates pixel color similarities, and the areas in the frame that ever experience changes while the vehicle rotates:</p>\n\n<p><img src=\"http://i.imgur.com/HjwOtxX.png\" alt=\"enter image description here\" title=\"\"></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "208038": "Hi,\n\n\"Currently the library offers 37 BGS algorithms.\" :\n\nhttps://github.com/myfavouritekk/bgslibrary\nhttps://github.com/andrewssobral/bgslibrary\nTrying to figure out if it can work easily **on still images.** \n\n\"A comprehensive review of background subtraction algorithms\":\n\nhttps://www.researchgate.net/publication/259340906_A_comprehensive_review_of_background_subtraction_algorithms_evaluated_with_synthetic_and_real_videos\n\nWould love to read more if you have something interesting. \n\nBest,",
    "208040": "Each vehicle has 16 frames of rotation, so within each set, the background should be static (save for some minor JPG noise). This may lend to techniques that require more frames. \n\nSome algorithms depend on a reference frame, and unfortunately we can't offer a reference frame without a vehicle, as such a photo is not taken for each vehicle. Although rare, there can be slight adjustments made to the camera, white balance/exposure (not sure if this is automatically optimized per-vehicle), and such over time.\n\nOne word of caution when approaching background subtraction, though, is that reflective objects can be problematic, as pixels on the object can end up being the same color as the background was when the object was not there.\n\nThis isn't an example of this effect specifically, but it illustrates pixel color similarities, and the areas in the frame that ever experience changes while the vehicle rotates:\n\n![enter image description here][1]\n\n  [1]: http://i.imgur.com/HjwOtxX.png"
  },
  "source": "meta"
}