Motion vectors and deep neural networks for video camera traps

Commercial camera traps are usually triggered by a Passive Infra-Red (PIR) motion sensor necessitating delay between triggering and the image being captured. This often seriously limits ability to record images of small fast moving animals. It also results in many “empty” images, e.g., owing foliage against background different temperature. In this paper we detail new mechanism based solely on sensor. is intended for use citizen scientists deployment an affordable, compact, low-power Raspberry Pi computer (RPi). Our system introduces video frame filtering pipeline consisting movement image-based processing. makes Machine Learning (ML) feasible live stream RPi. We describe our free open-source software implementation system; introduce suitable ecology efficiency measure that mediates specificity recall; provide ground-truth clip collection from traps; evaluate effectiveness thoroughly. Overall, trap turns out be robust effective.