Animal Pose Analysis Using Deep Neural Networks

Abstract

Animal pose tracking has historically been very time-consuming and inefficient. Fortunately, the new technology of deep learning networks is here to facilitate research. This incredible AI tool can greatly increase the efficiency of marking data and automatically parse them into desired categories. LEAP is a deep learning system based on CUDA neuro-network. With a short duration of training, LEAP can mark out subsequent data with very high accuracy by itself, which speaks great potential for analyzing long-time observational data. This paper aims to implement a deep learning system to analyze long-time data and validate the possibility of AI-assisted behavioral research.