OmniEye improves video and audio review for police body cameras
OmniEye: Efficient Multimodal Forensic Video Intelligence for Law-Enforcement Body-Worn Cameras
Emerging Technologies
Summary
Police officers often use body cameras that record video and sound, but reviewing all the footage can be slow and difficult. The authors created OmniEye, a system that watches video and listens to audio together in 30-second parts, then organizes the information so officers can quickly search and ask questions about what happened. This system runs efficiently on common hardware, making it easier for law enforcement to analyze footage during training or investigations.
What this means in practice
- •For law enforcement teams: Enable faster and detailed searches of body camera video and audio for training and review purposes using integrated multimodal understanding.
- •For security operations centers: Improve monitoring and analysis of surveillance footage by combining video and audio signals queried interactively with a joint multimodal model.
Authors
Mamadou K. Keita, Angela Srbinovska, Anita Srbinovska, Nishka Desai, Isabella Zicari, P. Kwaku Sanaah-Faried, Sanjay Charitesh Makam, Wyatt Auten, Vivek Senthil, Hannah Desnick, Jonathan Bateman, Adrian Martin, Christopher Homan, John McCluskey, Ernest Fokoué
Abstract
We introduce OmniEye, a multimodal video intelligence system for law-enforcement training and review (source code available on request to verified law-enforcement and public-safety agencies). OmniEye ingests body-worn camera footage and perceives every 30-second window jointly across video and audio with one multimodal foundation model. It then stores the model's structured output in an embedded SQLite database with BM25 full-text search. Officers can question the footage through an agent that writes structured queries, retrieves candidate windows, and re-perceives them with the model before it may cite them. The whole system runs on one 16 GB GPU with a 4-bit quantization-aware-trained model, and it also scales to full bf16 precision on a multi-GPU cluster.