TAIPEI (Taiwan News) — Taiwan and Russia have developed one of the fastest real-time object detection systems already in use to solve traffic flow problems in Taiwanese cities, the Academia Sinica announced Thursday (July 2).
“You Only Look Once” version 4 (YOLOv4) contributed to ending traffic jams in Taoyuan City and in Hsinchu City, according to a CNA report. The system’s accuracy had reached 43.5 percent, or 10 percent more than its previous version, the Academia Sinica said.
Elan Corporation (義隆電子) helped with the development, which took five months and was completed last April, according to officials. The YOLOv4 system only had to look at one image to understand the position of the objects, a talent which was useful not only for traffic control, but also for the development of self-driving vehicles, the analysis of medical images, and the testing of faulty equipment in factories.
YOLOv4 is one of the fastest object recognition system on the market, according to the key figures behind the project, Mark Liao (廖弘源) and Wang Chien-yao (王建堯) of the Academia Sinica and Russian developer Alexey Bochkovskiy.