Pranav Manjunath

Summary

For this project, the team aims to replicate an earlier work by Hussain et al. on comparing different augmentation techniques applied on a medical imaging dataset (DDSM) by their testing accuracy and ROC curves. Specifically, the team focuses on four augmentation techniques: Gaussian blur, Gaussian noise, Random rotations, Color jittering. The team shows that some augmentation techniques are able to extract medical image statistics more effectively than others, which will lead to better predictive performance in the corresponding models.