ML class project for cancer detection using deep learning.

A broad breakup of the tasks to be done :

1. Fetch data from here and focus only on the category of images that corresponds to lung cancer : https://stanfordmlgroup.github.io/projects/chexnext/

2. Train a classifier to maximize accuracy on this. Can liberally do transfer learning here and build off pretrained CNNs.

3. Grad-cam the output to highlight pixels that most strongly correspond to the output class : https://github.com/ramprs/grad-cam

4. Use this dataset to train Variational Autoencoders for data augmentation (a useful reference here : https://www.scitepress.org/Papers/2018/66186/66186.pdf)

5. Put everything together in a Github page, as asked for.

A couple of other useful references :

1. A link to the paper itself, which is a very extensive account of this very task (the first of the task, rather) : https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1002686

2. Link to the description of the dataset: https://arxiv.org/pdf/1901.07031.pdf

3. General PyTorch tutorials : https://pytorch.org/tutorials/