Research on Image Semantic Segmentation Model Based on Bidirectional Recursive Double Branch Network

Abstract

To solve the problem of road safety, express logistics industry, people begin to pay attention to no delivery, distribution and differences apparent change of scene scale is big, this paper presents a bidirectional recursive double branch network, reusable building blocks, in the form of circulation use of pairs of forward and backward jump connection to form the complex relationship between the encoder and decoder, at the same time, the double branch network is used to combine coarse information with fine information in the decoding stage, so as to restore the feature map more precisely, which helps to further improve the accuracy of the network.