Cattle breed identification and live weight evaluation on the basis of machine learning and computer vision

Abstract

The problem of the cow's live weight estimation is considered. A convolutional neural network based method for animal recognition and its breed identification in combination with epipolar geometry approach for object's size measurement is proposed. Information regarding animal's size and its breed is further used for LW estimation by multilayer perceptron based predictive model. This approach can be used to replace traditional methods of direct observation and measurement. The proposed system can be widely used in the management of a modern farm. Accuracy and performance of the proposed method has been tested with the participation of the experts.