A Practical Embedded System Utilizing Artificial Intelligence Technology

Thereis a clear and present lethal clash between South Africa's wildlife and its expanding population. Some of its main effects are harm to people and property. Fatalities, crop damage, livestock threads, and habitat devastation. We are inspired by the need for data-driven approaches in artificial intelligence in today's technologies. We have built an embedded system called “SmartFarmApp” by leveraging datasets for object detection. This system capably serves as the backbone of computer vision models by focusing on detecting and classifying livestock according to their classes using the Florence 2 AI (Artificial Intelligence) model. We have livestock of interest and wild animals as a threat to our livestock and if wild animals are around this class label the message for security concern must be raised. This system utilizes the Florence 2 AI vision-language model developed by Microsoft that excels in various vision tasks, including object detection where we use object detection Application Programmable Interfaces (APIs) to do detection for this system in a realistic outdoor environment. However, Florence 2 as vision tasks it does not support Object tracking, so we improved Florence 2 model to work by allowing tracking. Florence 2 model feeds our system, while built-in Raspberry Pi devices equipped with cameras feed this system in real-time using the mobile application to remotely view daily activities in our farms.