Research Article
Optimization of Loss Function for Pedestrian Detection
@INPROCEEDINGS{10.1007/978-3-030-97124-3_39, author={Shuo Zhang and Kailiang Zhang and Yuan An and Shuo Li and Yong Sun and Weiwei Liu and Likai Wang}, title={Optimization of Loss Function for Pedestrian Detection}, proceedings={Simulation Tools and Techniques. 13th EAI International Conference, SIMUtools 2021, Virtual Event, November 5-6, 2021, Proceedings}, proceedings_a={SIMUTOOLS}, year={2022}, month={3}, keywords={Computer vision Deep learning Pedestrian detection Loss function}, doi={10.1007/978-3-030-97124-3_39} }- Shuo Zhang
Kailiang Zhang
Yuan An
Shuo Li
Yong Sun
Weiwei Liu
Likai Wang
Year: 2022
Optimization of Loss Function for Pedestrian Detection
SIMUTOOLS
Springer
DOI: 10.1007/978-3-030-97124-3_39
Abstract
The advanced intelligent driving assistance system has improved the current traffic congestion to a great extent and effectively reduced frequent traffic safety accidents. Pedestrian detection technology is the core of autonomous driving technology, and its accuracy, real-time and complexity will directly determine the safe operation of autonomous driving. In the case of heavy traffic, detecting a single pedestrian in a crowd is still a challenging problem. Considering the problem of mutual occlusion between pedestrians in dense crowds, an improved function algorithm based on YOLOv3 is proposed to optimize the loss function and increase the accuracy of detection by replacing the anchor frame. Experimental results show that this method can effectively reduce the missed detection rate, increase the average accuracy, and help improve the effectiveness of pedestrian occlusion detection, ensure accurate pedestrian detection in traffic congestion scenarios, and ensure driving safety.
