Collection Aware Personalization (CAP) using Multi Modal Data Sources

This repository was built for the purpse of conducting research on content personalization in an image and text context. In contrast to using a pretrained generic model to achieve this, such as CLIP, we incorporate a personalization module that explicitly consumes a user's image library, and then updates an Image to Text model architecture to respond more effectively to a given user's text queries, and return user-conditional images from a large image library.