Datonymous AnonNet

De-identifing patient faces for facilitating protected health information (PHI) exchange among relevant stakeholders.

Team Members

• Abhi Sharma
• Jasim Alazzawi
• Josh Matz
• Prathamesh Prabhudesai

Abstract

This repository contains an implementation of AnonNet, our submission to the School of AI Hackathon on March 31st, 2019. AnonNet detects key identifying features on a person's face, i.e., the eyes, eyebrows, and mouth. Using this information, accurate pixelation and/or censor bar placement can de-identify an individual. This has several benefits, including but not limited to:

• Reducing the risk of protected patient information from being leaked
• Reducing the administrative burden of manually de-identifying patient information
• Increasing the availability and access to anonymized information, leading to innovation in the Healthcare space

Instructions to Reproduce

In order to replicate the AnonNet Jupyter Notebook, you'll need to install Node.js.

Acknowledgements

Thank you to the School of AI and Accenture for hosting this hackathon!