Traffic sign detection in unconstrained environment using improved YOLOv4

Open Webpage

Swastik Saxena, Somnath Dey, Miten Shah, Sundesh Gupta

Published: 01 Jan 2024, Last Modified: 06 Nov 2025Expert Syst. Appl. 2024EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•We have presented a model based on YOLOv4 for Traffic Sign Detection and Recognition.•Path Aggregation Network of YOLOv4 model is improved for better feature propagation.•We have used grouped convolutional layers to reduce the number of parameters.•A new dataset of Indian traffic signs is proposed with more than 3000 images.•Cross-dataset experiments are conducted on GTSDB and Indian traffic signs.