Google CAHSI Project (Video Anonymization Pipeline)

As part of a grant from Google Research, Cal Poly and UCR began an joint effort to discover whether LLMs could perform better at anomoly detection than a traditional lightweight classifier. This required a video anonymization pipeline to be built and deployed to Delta, the HPC funded by ACCESS credits provided by NSF, as non-CPP students are not permitted to view data that may contain personally identifiable information (PII) of patients.

From October to December 2025, two major versions were developed which can be found here . This repo contains models trained on patient data as well as YOLO training summaries which include galleries of model outputs overlayed on patient stills. As such, this repo is private, and you may request access from tingtingchen@cpp.edu. A technical report of the project can be found here .