Semantic segmentation method in automatic driving scene, storage medium and electronic equipment

The invention relates to a semantic segmentation method in an automatic driving scene, a storage medium and an electronic device, and the method comprises the steps: obtaining a road color image of an automobile visual angle, and carrying out the preprocessing of the image to obtain an image X; obtaining semantic features of different scales of the image X based on an encoder of a mixed transformer; extracting spatial information of semantic features of different scales of the image X by using graph convolution, and obtaining spatial semantic features of each scale through node message transmission; a hypergraph concept is used, a graph structure of each scale is regarded as a hypergraph, hyperedge representation of the hypergraph is solved, interaction of information of different scales is achieved through hyperedges, and multi-scale space mixed semantic information is obtained; a segmentation result is obtained by mixing decoders and fusing multi-scale spatial semantic information, multi-scale global semant