{
  "id": 52662,
  "title": "Mixed-Scale Dense CNN for segmentation?",
  "url": "/competitions/data-science-bowl-2018/discussion/52662",
  "author_name": "",
  "post_date": "2018-03-21T20:40:06.392083Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I saw the following paper recently released (Dec 2017) discussing a novel type of model to perform segmentation:\n<a href=\"https://slidecam-camera.lbl.gov/static/asset/PNAS.pdf\">https://slidecam-camera.lbl.gov/static/asset/PNAS.pdf</a></p>\n\n<p>Has anyone tried implementing a Mixed-Scale Dense CNN for this challenge? Their paper seems to show better performance at segmentation than UNet, so would be interested to hear if anyone has actually tried implementing this here. The paper also says they have far fewer parameters than UNet models, and suggests their model can train better on fewer input images as a result - which could be interesting for this competition since there are relatively few training images.</p>",
  "messages": [
    {
      "id": "300696",
      "postDate": "03/21/2018 20:40:06",
      "content": "<p>I saw the following paper recently released (Dec 2017) discussing a novel type of model to perform segmentation:\n<a href=\"https://slidecam-camera.lbl.gov/static/asset/PNAS.pdf\">https://slidecam-camera.lbl.gov/static/asset/PNAS.pdf</a></p>\n\n<p>Has anyone tried implementing a Mixed-Scale Dense CNN for this challenge? Their paper seems to show better performance at segmentation than UNet, so would be interested to hear if anyone has actually tried implementing this here. The paper also says they have far fewer parameters than UNet models, and suggests their model can train better on fewer input images as a result - which could be interesting for this competition since there are relatively few training images.</p>",
      "rawMarkdown": "I saw the following paper recently released (Dec 2017) discussing a novel type of model to perform segmentation:\nhttps://slidecam-camera.lbl.gov/static/asset/PNAS.pdf\n\nHas anyone tried implementing a Mixed-Scale Dense CNN for this challenge? Their paper seems to show better performance at segmentation than UNet, so would be interested to hear if anyone has actually tried implementing this here. The paper also says they have far fewer parameters than UNet models, and suggests their model can train better on fewer input images as a result - which could be interesting for this competition since there are relatively few training images.",
      "votes": null
    },
    {
      "id": "300903",
      "postDate": "03/22/2018 03:13:50",
      "content": "<p>Interesting, I will take a look at that and I will reply. \nThanks!</p>",
      "rawMarkdown": "Interesting, I will take a look at that and I will reply. \nThanks!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 300903,
      "author_name": "coronate",
      "author_url": "",
      "post_date": "03/22/2018 03:13:50",
      "content": "<p>Interesting, I will take a look at that and I will reply. \nThanks!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "300696": "I saw the following paper recently released (Dec 2017) discussing a novel type of model to perform segmentation:\nhttps://slidecam-camera.lbl.gov/static/asset/PNAS.pdf\n\nHas anyone tried implementing a Mixed-Scale Dense CNN for this challenge? Their paper seems to show better performance at segmentation than UNet, so would be interested to hear if anyone has actually tried implementing this here. The paper also says they have far fewer parameters than UNet models, and suggests their model can train better on fewer input images as a result - which could be interesting for this competition since there are relatively few training images.",
    "300903": "Interesting, I will take a look at that and I will reply. \nThanks!"
  },
  "source": "meta"
}