Moving Image YOLOv3 Process Comparison with Multiple Resolutions and Frames-per-Second
Moh Noor Al-Azam, Mochamad Mizanul Achlaq, Cahyo Darujati · 2022
Many city/district governments have installed Closed-Circuit Television (CCTV) on the streets as a form of their smart city. This CCTV installation is usually only useful as a recording of events at a location that needs to be seen at a time. Actually, with the same capital, moving image recordings produced by CCTV can also be used for various things, such as vehicle counters that can show the level of road density. In this paper, the researcher tries to use “You Only Look Once” or YOLO with several configurations of the moving image feature to recognize objects as vehicles. Because one factor that need to be considered in object detection for moving images is the need for large computational resources and long processing time. This study examines the effect of reducing the resolution size feature and the number of frames per second (fps) of moving images in the YOLO process, to obtain optimal use of computing resources and processing time. Reducing the resolution and the number of frames per second does not have much effect on the use of Central Processing Unit (CPU) computing resources used by researchers, which is around 9S.5l% on average (with Xeon 8 cores). However, it can reduce processing time by up to 70%.