An improved PSPNet-Based Semantic Segmentation Method for Urban Street Scenes
In the current process of streetscape semantic segmentation task,the traditional semantic segmentation methods are prone to the problems of imprecise edge segmentation and serious influence of irrelevant background fea-tures.The model consists of a semantic segmentation sub-network and an edge detection sub-network,in which an attention mechanism is embedded in the semantic segmentation sub-network to enhance the acquisition of effective features and ignore irrelevant feature information,and the edge detection sub-network is used to obtain more accurate contour features.Finally,the final result is obtained by fusing the two features through the feature fusion module.Some ablation experiments were conducted on the Cityscapes dataset,and the model improved the average delivery by 3.3％ compared to the original model of the comparison,and the experimental comparison with the existing model,and the experimental results proved that the E-PSPNet model can effectively improve the problem of imprecise edge seg-mentation of streetscape and the serious influence of background irrelevant features.
Semantic segmentationEdge detectionConvolutional block attention moduleCity street view
NETL
NSTL
万方数据
