Dense Pixel-level Beef Cattle Instance Segmentation Using a Fully Convolutional Network

Aram Ter-Sarkisov, Robert Ross, John Kelleher, Bernadette Earley, Michael Keane

Abstract:

We present an instance segmentation algorithm trained and applied to a video of a group of heifers recorded in a winter finishing feedlot. We transform a fully convolutional class segmentation network into an instance segmentation network that
learns to label each instance of a heifer separately. We introduce two new modules that teach the network to output a single prediction for every animal. These results are an early contribution towards behaviour analysis in winter finishing beef cattle
for early detection of welfare-related problems, namely lameness.

Keywords: deep learning, instance segmentation, precision farming, convolutional neural networks

TL;DR: Instance segmentation algorithm to identify beef cattle in a winter finishing feedlot facility