Vehicle Detection Based on Improved YOLO-LITE Algorithm
Download as PDF
DOI: 10.23977/iset.2019.040
Author(s)
Shuai Zhao, Fucheng You and Shaomei Wang
Corresponding Author
Shuai Zhao
ABSTRACT
With the increasing traffic congestion in urban roads, people have higher and higher requirements for real-time monitoring of vehicles, but the traditional vehicle detection algorithms are too demanding on computer hardware. Like the YOLOv3 algorithm, although the performance aspect is more objective, it requires high hardware equipment. Therefore, based on such problems, the YOLO-LITE algorithm is proposed. This algorithm achieves the goal of introducing vehicle detection into a non-GPU computer, and experiments have shown that the YOLO-LITE algorithm is not satisfactory in terms of mAP, but the FPS aspect is very good, and the YOLO-LITE algorithm is 8.8 times faster than Tiny-YOLO algorithm.it is very helpful for the lack of GPU computers to achieve vehicle detection.
KEYWORDS
YOLO-LITE, YOLO, deep learning, vehicle detection