slideflow VS stylegan2-slideflow

Easy-to-use API with clear documentation

I run an open source, medical AI project for digital pathology called Slideflow at the University of Chicago. We’re working on developing reliable biomarkers for patients with lung, breast, and thyroid cancer, and we can always use more help! Got lots of interesting active projects - uncertainty quantification, generative models, embedded systems deployment - send me a DM if it sounds interesting!

After years of development, we've released our open-source deep learning framework for digital histology, Slideflow (https://github.com/jamesdolezal/slideflow). It has flexible and highly optimized whole-slide image processing, support for a wide variety of existing and custom architectures (with continuous, categorical, or time-series outcomes), real-time digital stain normalization, a number of explainability tools, and integrated uncertainty quantification. It's compatible with both Tensorflow and PyTorch, available on PyPI and DockerHub, and comes with good documentation (https://slideflow.dev/). We've tried out a number of alternative frameworks over the years, and I think the easy of use, flexibility, and performance optimizations set it apart from other repos you'll find on GitHub.

We have! I have a slightly modified version of StyleGAN2 published to enable easy interfacing with our pipeline (https://github.com/jamesdolezal/stylegan2-slideflow). One of our recent prepints used generative models, and in that manuscript's associated code, we included a short section explaining how to train StyleGAN2 with Slideflow: https://github.com/jamesdolezal/biscuit#gan-training