Real-time semantic segmentation based on attention mechanism and multi-branch feature fusion
In order to balance between accuracy and real-time performance in semantic segmentation,based on the fast convolutional neural network model(Fast-SCNN),this paper proposes a real-time semantic segmentation algorithm model which combines the attention mechanism and the multi-branch feature fusion.First,the attention module captures the interrelation between spatial features to enhance the spatial details.Second,the fusion module is designed to maximize the information of each branch to achieve a better fusion of deep features and shallow features.Finally,an adaptive feature enhancement attention module is introduced to capture the interdependencies between long distance pixels.The experimental results show that the proposed algorithm achieves 71.55%segmentation accuracy on Cityscapes dataset,a reasoning speed FPS of 97.6 frame/s,and the parameter number of of 1.39 M.These demonstrate the effectiveness of the network model constructed by the algorithm.
万方数据
维普
