Research on urban street view semantic segmentation algorithm based on DeepLabv3+algorithm
In the field of autonomous driving,semantic segmentation of urban street scenes is crucial for improving the safety and efficiency of the system.Aiming at the problems of too many parameters,insufficient portability and limited segmentation effect of traditional semantic segmentation models,this paper proposes a solution based on improved DeepLabv3+.This improved model incorporates a lightweight MobileNetv2 backbone network and the SE attention mechanism,and optimizes the Atrous spatial pyramid pooling(ASPP)module from a parallel to a serial structure with a depth-separable convolutional structure.On the Cityscapes dataset,the method in this paper achieves 75.90%MIoU,which significantly improves the segmentation accuracy and computational efficiency.
NETL
NSTL
万方数据
