StreamPPG: Low-Latency rPPG Estimation via Consistent Privileged Learning
StreamPPG is a new architecture for remote photoplethysmography (rPPG) that estimates blood volume pulses from facial videos in real-time, frame by frame. It solves the problem of high latency in existing methods by using a consistent privileged learning strategy and an adaptive temporal modeling module to maintain accuracy while processing video streams immediately.