Deep learning based animal recognition system for feeding system

Traditional animal feeding methods, due to their reliance on manual practices, fall short in meeting the demands of modern animal husbandry. This leads to issues such as feed wastage, irregular feeding cycles, and unauthorized animal access to feeders. Specifically, uncontrolled feed consumption results in both economic losses and poses significant risks regarding hygiene and security. This study aims to implement deep learning-based image processing techniques to enable more efficient and controlled management of animal feeding processes. Within the scope of this study, YOLOv8, YOLOv9, YOLOv11, and the developing YOLOv12 algorithms were used for the detection of bird, cat, and dog species. These models allow for fast and accurate animal recognition thanks to their real-time object detection capabilities. Furthermore, the classification performance was significantly enhanced by utilizing Convolutional Neural Network (CNN)-based deep learning architectures such as ResNet-50, MobileNet, and EfficientNetB0 during the recognition process. Based on the data obtained from images, only identified and authorized animals are permitted access to feeding areas, thereby controlling feed consumption, preventing resource waste, and maintaining hygienic conditions. The proposed artificial intelligence-based approach has a wide range of applications, from animal shelters to rural livestock farming areas. Future work aims to diversify the animal species that can be recognized, improve training processes with more comprehensive and balanced datasets, ensure integration with low-power embedded hardware, and develop an interface offering remote access via mobile applications. This study demonstrates the effectiveness of AI-powered solutions in animal recognition-based feeding applications, offering a sustainable and secure framework.