Animal behavior classification via deep learning on embedded systems

Reza Arablouei, Liang Wang, Lachlan Currie, Jordan Yates, Flavio A. P. Alvarenga, Gregory Bishop-Hurley

Abstract: Highlights•We develop a new algorithm to classify animal behavior in situ and in real time.•It performs feature extraction and classification via end-to-end deep learning.•We evaluate its performance via real-world datasets collected from grazing cattle.•It outperforms the state-of-the-art in both classification accuracy and complexity.•Its inference is realized efficiently on the embedded system of a smart collar tag.