CLIP-based Histology Annotation Predictor

We attempt to automate medical diagnosis by using the CLIP framework to create an image-captioning tool that retrieves useful textual pathology notes given histology images. We implement a modular CLIP framework, experiment with candidate encoders, and demonstrated the ability to identify the pathology of tissue images from unstructured data.