Performance Comparison of Yolo-Lite and YoloV3 Using Raspberry Pi and MotionEyeOS

This paper proposes system comparison on identifying and processing of human image based on YOLOLITE and YOLOV3 algorithms. Computer Vision (CV) is a field of computer science where the focus is on learning how computers can be trained to identify and process image data as humans do. There are many open source CV frameworks have been proposed such as OpenCV. This paper shows a comparison between YOLO-LITE and YOLOV3 algorithms and analyzes their performance. We have implemented both algorithms in several test cases in the real time domain and carried out in the same test environment. The result shows that the Raspberry Pi camera worked at 15 fps on YOLO-LITE and 1 fps on YOLOV3. This indicates that YOLO-LITE has an average performance of 1 second faster while YOLOV3 has an average accuracy of 30% better.