PhD Forum Abstract: Activity Classification at the Edge

This study investigates activity classification using deep learning for developing an energy efficient and high-performance behavioural model on edge (resource constrained) devices. The specific focus of this research is on cattle activity classification. Considering accuracy and efficiency of approaches, different deep neural network structures and data representations will be explored. The developed model will be capable of identifying multiple cattle behavioural patterns accurately. However, there are some challenges to develop energy efficient and computationally light systems.