Automatic Rodent Detection from Motion-triggered Wildlife Cameras

Abstract

Active monitoring is indispensable in containing the spread of contagious diseases among animal stocks, particularly in detecting any intruders that may introduce harmful pathogens into livestock farms. Given the time-consuming nature of video monitoring and human attention limitations, there is a need for autonomous systems capable of continuously screening videos from the moment the camera starts recording. In this thesis, we utilize Deep Learning object detectors coupled with object trackers which can significantly automate these tasks reducing human intervention.