In this project, we would train a StyleGAN2 model for medical datasets. This could be beneficial for synthetic data augmentation, or potentially encoding into and studying the latent space could be useful for other medical applications.
Datasets
Personally, I am more interested in histopathological datasets:
Hey @tmabraham, super interested in being part of this project. I’m working in the intersection on Deep Learning and Medical imaging and have worked on histopathological datasets.
Really looking forward to train StyleGAN2 using JAX and see how the latent space on medical imaging pans out for other downstream tasks.
It should definitely be feasible to write a training script during the week! Let’s officially create this project GANs are usually quite difficult / unstable to train, but let’s give it a shot!
I’m a Master’s student from India.
This is a very interesting project. I have some experience in working with GANs and used them to generate synthetic data (agricultural sector).
I am unfamiliar with JAX. This will a great opportunity for me to learn and contribute. Looking forward to working on this project.
Hey @ tmabraham ! This project sounds interesting! I would love to join it. Although I have not worked with JAX yet, I have experience with GANs and Computer Vision using PyTorch. It would be a great learning experience for me. Please let me know how can I join?
Hi @patrickvonplaten@tmabraham I have been working on digital pathology for my masters, and I feel I can contribute to this project. I have experience with GANs too. Please add me to the project sheet if there are still spots left
Are you still interested? Please contact Patrick Von Platen (over here or on Slack) regarding this. Also I’ll need your Discord username, that’s where we are discussing.