How Reveal GPU Redaction Works
The Reveal team gathered insights from various law enforcement departments to address common challenges of video redaction.
"Redacting video, especially from body cameras, is a critical yet cumbersome task that often overloads our systems," explains Oliver Gibbs, Head of Product Management at Reveal. "Our customers needed a more robust solution to handle extensive video footage without sacrificing system performance, so we implemented our GPU Redaction service to meet these demands."
The newly enhanced GPU Redaction operates on a dedicated Azure GPU Virtual Machine (X64), selected specifically to boost processing speeds and protect the integrity of DEMS 360's core functionalities. This setup supports heavy loads, such as large volumes of body camera footage, and offered the opportunity to implement some processing enhancements.
How it works:
- Automatic Face Detection: Once footage is uploaded, the enhanced face detection technology scans each frame more effectively, increasing the accuracy of face identification. This improvement ensures that individuals are consistently recognised and tracked throughout the video.
- Sequence Grouping and Tracking: Faces identified are not only detected but are also grouped into sequences. The software uses sophisticated facial recognition supported by spatial logic to track these sequences across frames, maintaining continuity even when faces move or temporarily leave the frame.
- Extended Redaction Frames: To ensure complete privacy, the system automatically extends the redaction to six frames before and after a face is detected. This feature minimises the risk of missing any part of the sequence where the individual is visible.
- Customisable Redaction Borders: Administrators can set specific redaction parameters, such as the size of the border around detected faces. These settings are consistently applied across the Redactor user interface, ensuring uniformity in the redaction throughout the video.
Exporting the redacted footage is straightforward. Users can simply select the export option, choose the desired file format, and distribute the redacted video securely. Original footage remains secure and unaltered in the system, ensuring that data integrity is maintained throughout the redaction process.