Video Anonymizer for Hospital Research
AI/ML Consultant — AutomateThis
Unblocked research stalled on ethics approval
- Generative Adversarial Networks
- Computer Vision
- Model training
- Model hosting
- Video processing
Problem
UHN had recorded video needed for a study, but couldn't get research ethics board approval to use it — patients and staff were identifiable on camera, and de-identifying footage by hand wasn't practical at the volume required. The research was stalled, not because the science was hard, but because the privacy problem hadn't been solved.
Approach
Built a video anonymization tool that detects and obscures identifying features — faces and bodies — frame by frame, tuned to hold up under ethics review rather than just "look blurred." The bar wasn't visual anonymization for its own sake; it was producing footage a research ethics board would actually sign off on.
Outcome
The anonymized footage cleared ethics review, unblocking research that had been stuck waiting on a privacy solution rather than on the underlying science.