This code is the companion for the tutorial located at https://www.kaggle.com/c/data-science-bowl-2017#tutorial

This repository is not intended to be an out of the box solution for the DSB challenge. It will not run out-of-the-box without editing. That was not it's intention. The tutorial was put together rapidly by several people working in tandem and the code herein is a collection of the code they used to produce the tutorial found on the DSB website.

The intent behind this tutorial was to presented a series of steps that can be followed as a starting point for competitors. Our hope is that this can save competitors time in framing the problem and that they can lift some of this code to speed up their own solution generation. We expect that the competitors efforst will supercede this tutorial in short order--which is, of course, the point of the competition.

Thanks for participating and helping to advance cancer diagnosis!

#FAQ

The unet weights supplied here were generated by training on a fraction of the LUNA 2016 data and the training set preparation could be improved on. Some nodules are readily identified by the unet weights provided here, and others are not. We trained enough that it seemed that in principle this is a reasonable approach. If you would like to initialize your u-net with these weights, or use these to test out how a u-net segmentor works, you might find these weights useful.

Please read the "About this repository" section

If you would like to submit a pull request to fix an issue that you found, we'll be happy to review it. Otherwise, we may not get around addressing any errors you might find in this code. Making a note in the issues section will be very helpful for other people that are perusing the repo.