{
  "id": 117489,
  "title": "Analysis of the 2018 Data Science Bowl is now published!",
  "url": "/competitions/data-science-bowl-2018/discussion/117489",
  "author_name": "Anne Carpenter",
  "post_date": "2019-11-15T21:50:24.152000",
  "votes": 5,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I am happy to report our paper analyzing the DSB is now online at Nature Methods! (<a href=\"https://www.ncbi.nlm.nih.gov/pubmed/31636459\">here</a>) (<a href=\"https://personal.broadinstitute.org/anne/publications/127_Caicedo_NatMethods_2019.pdf\">PDF here</a>). All authors are members of my lab except we also invited CherKeng Heng, to contribute based on his many helpful comments during the competition.</p>\n\n<p>The top three solutions were very different from each other and offered unique insights to the problem, especially in pre/post processing techniques. Overall, we conclude that the challenge was successful: top models can instantly segment an image from an unseen experiment with no parameter tweaking or fine tuning or image annotations to a level that exceeds an expert quickly tuning a classical image processing pipeline. There is even a user-friendly web interface for one of the models stemming from this competition (<a href=\"https://www.nucleaizer.org\">https://www.nucleaizer.org</a>) and already we are seeing many computer vision papers citing use of the annotated images from this competition, a tremendous resource for the community!</p>\n\n<p>I've updated the post on <a href=\"https://www.kaggle.com/c/data-science-bowl-2018/discussion/47864#latest-314862\">how to cite</a>.</p>\n\n<p>Thanks to all who made this a great competition!</p>",
  "messages": [
    {
      "id": 674080,
      "postDate": "2019-11-15T21:50:24.153Z",
      "content": "<p>I am happy to report our paper analyzing the DSB is now online at Nature Methods! (<a href=\"https://www.ncbi.nlm.nih.gov/pubmed/31636459\">here</a>) (<a href=\"https://personal.broadinstitute.org/anne/publications/127_Caicedo_NatMethods_2019.pdf\">PDF here</a>). All authors are members of my lab except we also invited CherKeng Heng, to contribute based on his many helpful comments during the competition.</p>\n\n<p>The top three solutions were very different from each other and offered unique insights to the problem, especially in pre/post processing techniques. Overall, we conclude that the challenge was successful: top models can instantly segment an image from an unseen experiment with no parameter tweaking or fine tuning or image annotations to a level that exceeds an expert quickly tuning a classical image processing pipeline. There is even a user-friendly web interface for one of the models stemming from this competition (<a href=\"https://www.nucleaizer.org\">https://www.nucleaizer.org</a>) and already we are seeing many computer vision papers citing use of the annotated images from this competition, a tremendous resource for the community!</p>\n\n<p>I've updated the post on <a href=\"https://www.kaggle.com/c/data-science-bowl-2018/discussion/47864#latest-314862\">how to cite</a>.</p>\n\n<p>Thanks to all who made this a great competition!</p>",
      "rawMarkdown": "I am happy to report our paper analyzing the DSB is now online at Nature Methods! ([here](https://www.ncbi.nlm.nih.gov/pubmed/31636459)) ([PDF here](https://personal.broadinstitute.org/anne/publications/127_Caicedo_NatMethods_2019.pdf)). All authors are members of my lab except we also invited CherKeng Heng, to contribute based on his many helpful comments during the competition.\n\nThe top three solutions were very different from each other and offered unique insights to the problem, especially in pre/post processing techniques. Overall, we conclude that the challenge was successful: top models can instantly segment an image from an unseen experiment with no parameter tweaking or fine tuning or image annotations to a level that exceeds an expert quickly tuning a classical image processing pipeline. There is even a user-friendly web interface for one of the models stemming from this competition (https://www.nucleaizer.org) and already we are seeing many computer vision papers citing use of the annotated images from this competition, a tremendous resource for the community!\n\nI've updated the post on [how to cite](https://www.kaggle.com/c/data-science-bowl-2018/discussion/47864#latest-314862).\n\nThanks to all who made this a great competition!",
      "votes": 5
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "674080": "I am happy to report our paper analyzing the DSB is now online at Nature Methods! ([here](https://www.ncbi.nlm.nih.gov/pubmed/31636459)) ([PDF here](https://personal.broadinstitute.org/anne/publications/127_Caicedo_NatMethods_2019.pdf)). All authors are members of my lab except we also invited CherKeng Heng, to contribute based on his many helpful comments during the competition.\n\nThe top three solutions were very different from each other and offered unique insights to the problem, especially in pre/post processing techniques. Overall, we conclude that the challenge was successful: top models can instantly segment an image from an unseen experiment with no parameter tweaking or fine tuning or image annotations to a level that exceeds an expert quickly tuning a classical image processing pipeline. There is even a user-friendly web interface for one of the models stemming from this competition (https://www.nucleaizer.org) and already we are seeing many computer vision papers citing use of the annotated images from this competition, a tremendous resource for the community!\n\nI've updated the post on [how to cite](https://www.kaggle.com/c/data-science-bowl-2018/discussion/47864#latest-314862).\n\nThanks to all who made this a great competition!"
  }
}