Automatic image process weight estimation of sheep

In animal husbandry, monitoring the weight gains of livestock is an important issue. Changes in the weight of animals can be a sign of disease. Farm owners often try to weigh their animals with traditional weighing systems. This is a costly, difficult and laborious process. The aim of this study is to develop an automatic weight determination system based on image processing. For this purpose, a camera system that takes images of small cattle from three directions (top, side and back) as they leave the pen was created and the images were recorded. Analyses were made on the recorded images with the deep learning approaches (MobileNetV2 and Mask R-CNN) and the weights were tried to be estimated. The performance of the system was tested by comparing the actual weights of the animals taken with the classical weighing system with the weight estimate values obtained with the image analysis.