{
  "id": 303234,
  "title": "COTSBOT was getting 99.4% accuracy.",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/303234",
  "author_name": "",
  "post_date": "2022-01-26T15:06:25.433718900Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>According to this video, the original COTSBOT with ML was getting about 99.4% accuracy after being trained on images, and youtube videos.<br>\n<a href=\"https://www.youtube.com/watch?v=WXmvOjiYdvo\" target=\"_blank\">https://www.youtube.com/watch?v=WXmvOjiYdvo</a><br>\nCOTSBOT discussion is about the 5 minute mark in the video.</p>",
  "messages": [
    {
      "id": "1665095",
      "postDate": "01/26/2022 15:06:25",
      "content": "<p>According to this video, the original COTSBOT with ML was getting about 99.4% accuracy after being trained on images, and youtube videos.<br>\n<a href=\"https://www.youtube.com/watch?v=WXmvOjiYdvo\" target=\"_blank\">https://www.youtube.com/watch?v=WXmvOjiYdvo</a><br>\nCOTSBOT discussion is about the 5 minute mark in the video.</p>",
      "rawMarkdown": "According to this video, the original COTSBOT with ML was getting about 99.4% accuracy after being trained on images, and youtube videos.\nhttps://www.youtube.com/watch?v=WXmvOjiYdvo\nCOTSBOT discussion is about the 5 minute mark in the video.",
      "votes": null
    },
    {
      "id": "1665361",
      "postDate": "01/26/2022 19:05:17",
      "content": "<p>It looks like we've been given a different style of footage with a goal to perform population counts with fast moving, blurry diving footage rather than slow moving single target cots detection. I don't think anyone will reach 99+ acc especially with the occasional cots ignored due to visibility or cropped by edge of frame in the training data.</p>",
      "rawMarkdown": "It looks like we've been given a different style of footage with a goal to perform population counts with fast moving, blurry diving footage rather than slow moving single target cots detection. I don't think anyone will reach 99+ acc especially with the occasional cots ignored due to visibility or cropped by edge of frame in the training data.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1665361,
      "author_name": "maxvandijck",
      "author_url": "",
      "post_date": "01/26/2022 19:05:17",
      "content": "<p>It looks like we've been given a different style of footage with a goal to perform population counts with fast moving, blurry diving footage rather than slow moving single target cots detection. I don't think anyone will reach 99+ acc especially with the occasional cots ignored due to visibility or cropped by edge of frame in the training data.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1665095": "According to this video, the original COTSBOT with ML was getting about 99.4% accuracy after being trained on images, and youtube videos.\nhttps://www.youtube.com/watch?v=WXmvOjiYdvo\nCOTSBOT discussion is about the 5 minute mark in the video.",
    "1665361": "It looks like we've been given a different style of footage with a goal to perform population counts with fast moving, blurry diving footage rather than slow moving single target cots detection. I don't think anyone will reach 99+ acc especially with the occasional cots ignored due to visibility or cropped by edge of frame in the training data."
  },
  "source": "meta"
}