Lightweight Traffic Sign Recognition Algorithm based on Cascaded CNN
Sungick Kong, Jong-Hee Park, Sang-Seol Lee, Sung‐Joon Jang · 2019
Autonomous vehicle technology is evolving with deep learning. Traffic sign recognition informs the driver of necessary information when the driver does not recognize the traffic sign while driving or when the traffic sign information is missing from the GPS database. In this paper, we collected Traffic signs in South Korea and we proposed a light-weight traffic sign recognition (TSR) algorithm based on cascaded CNN. This algorithm is hardware-friendly and reduces the computational complexity and the number of computations compared to the previously announced YOLO v2-tiny. Our Korean traffic sign dataset was used to learn the network and verify the algorithm. Through this process, we have studied the future improvement plan to make algorithm that can be used for actual road driving.