Universal bovine identification via depth data and deep metric learning

Download PDF
Open Webpage

Asheesh Sharma, Lucy Randewich, William Andrew, Sion L. Hannuna, Neill W. Campbell, Siobhan Mullan, Andrew W. Dowsey, Melvyn L. Smith, Mark F. Hansen, Tilo Burghardt

Published: 01 Jan 2025, Last Modified: 07 Nov 2025Comput. Electron. Agric. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Breed-agnostic cattle ID system using depth maps and point clouds.•ResNet and PointNet achieve accuracies on par with the coat pattern-based method.•PointNet maintains accuracy despite a reduction in the number of input points.•Provides CowDepth2023 dataset with 21,000 depth-colour image pairs, with code.•Paves the way for real-time, AI-assisted cattle welfare systems.•Enables integration with activities like body condition for personalised care.