In this paper, we present a deep learning model for high-accuracy, high-speed detection of vehicle taillights in traffic. The model consists of three major modules: the lane detector, the car detector, and the taillight detector. Unlike most previousl...
In this paper, we present a deep learning model for high-accuracy, high-speed detection of vehicle taillights in traffic. The model consists of three major modules: the lane detector, the car detector, and the taillight detector. Unlike most previously proposed algorithms where hand-coded schemes are used, we have adopted a data-driven approach. Both the pipelined approach and data-driven approach are necessary since we need to deal with both false positives and detection misses without complex hand-crafted logic. Two different implementations are introduced in this paper. In the first implementation, while lane detection was performed using hand-crafted algorithm, we used the ResNet-RRC as the deep neural network for car and taillight detection. In the second implementation, the PINet was used for lane detection, and the YOLOv7 was used for car and taillight detection. The robustness of our model was verified using datasets from Sungkyunkwan University (SKKU) as well as the Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI). Our model works well even in hostile conditions. It achieves detection rates as high as 100% in testing with the SKKU dataset. When using the KITTI 2D Object dataset, the model achieves a taillight detection rate of 96%. The model achieves 100% taillight detection rate on a certain, small subset of the KITTI Tracking dataset. The system achieves real-time speeds not only on large-scale computers but also on embedded machines such as NVIDIA® Jetson™ Orin.