In CS156b, we were tasked with creating a model that could diagnose several pathologies in the CheXpert Dataset.
The project utilizes pytorch lightning to optimize training on multiple nodes in the Caltech HPC. We chose a pretrained vision transformer by facebook called DinoVIT for the base of our model. We also explored using a GAN to create more data and a segmentation model.
By Zack Dugue, Benja Share, Eli Kugelsk, Saraswati Soedarmadji, and Lucas Ancieta