Automated Cattle Classification and Counting Using Hybridized Mask R-CNN and YOLOv3 Algorithms

Computer vision is an interdisciplinary research topic that investigates how computers can understand image data or videos at a high level. One of the most promising but difficult problems in intelligent livestock management is accurate and reliable animal counts in camera-acquired imagery. To detect and count cattle, most contemporary systems rely on hardware such as sensors, drones, and algorithms such as CNN (Convolutional Neural Networks), R-CNN (Regional-Convolutional Neural Networks), and YOLO (You Look Only Once). Mask R-CNN has a greater prediction accuracy, but YOLOv3 efficiently classifies at a high speed, and these algorithms are often regarded as more accurate and quick methods. This work proposes a novel technique called Mask-YOLOv3 for classifying and counting mixed livestock, including cow, sheep, and horse, that combines the advantageous aspects of both of these algorithms in order to increase detection and counting accuracy. The proposed algorithm can be used effectively for early detection of undesired activities like cattle loss, cattle theft etc.