 This paper proposes a novel approach to dynamic machine vision which uses a photo memory stirrer embedded within a camera to recognize past and predict future movements of objects. By embedding the past frames into the current frame, the photo memory stirrer enables real-time perception of moving objects without requiring any additional data flow. This allows for more efficient and accurate dynamic machine vision applications such as autonomous driving and robotics. This article was authored by Hongwei Tan and Sebastian van Dijkken.