3D Geometric Shape Recognition System using YOLO v8 Implemented on Raspberry Pi

Abstract

: Recently, there has been a growing trend in the popularity of robots capable of performing complex tasks without human intervention, such as differentiating and

picking up products with various shapes. These robots have found applications in diverse fields like manufacturing and servicing. To enable object recognition, artificial intelligence (AI) technology is essential for the robots. However, implementing AI algorithms is challenging and requires significant computational power. Additionally, for mobile robots, the use of microcontrollers is not an optimal

choice; embedded processors are preferred. This paper introduces a system for recognizing 3D geometric shapes, employing YOLO v8 implemented on a Raspberry Pi. The proposed system combines computer vision techniques, a webcam, and a Raspberry Pi to identify two classes of 3D geometric shapes: cubes and spheres. The

experimental results demonstrate that the system achieves an impressive mean average precision (mAP) of 98.3%, precision of 97.3% and recall of 94.4%. It is worth noting that the system is cost-effective, easily integrated into existing setups due to its compact form factor, high mobility, and low power consumption. In summary, the proposed 3D geometric shape recognition system demonstrates the potential of employing YOLO v8 on Raspberry Pi for real-time, affordable, and highly accurate 3D object recognition.