Face blurring
Face blurring is the irreversible process of obscuring faces on the APU at the moment of capture, before any image can leave the device.
Face blurring is the irreversible process of obscuring faces on the APU at the moment of capture, before any image can leave the device. The blurring uses a one-way algorithm — the original face cannot be reconstructed from the blurred output, and the underlying raw frame is immediately discarded.
How it works
Face detection and blurring run as the first stage of the on-device pipeline. By the time any frame is held for more than a few milliseconds, faces have already been anonymised. This applies both to operational processing and to the short snapshot windows used during per-store AI tuning, where blurred images are temporarily retained so the Computer Vision team can verify staff recognition and demographic accuracy.
Because blurring happens at source and cannot be reversed, the system produces no biometric identifiers, no face templates and nothing that could re-identify a customer.