Context-aware embedded crowd counting method and system, medium and electronic equipment

The invention provides a context-aware embedded crowd counting method, a context-aware embedded crowd counting system, a medium and electronic equipment, wherein the context-aware embedded network ismainly used for extracting low-level spatial features of an image, segmenting the image into super-pixels, fusing a result of segmenting the image into the super-pixels and an extraction result of thelow-level spatial features, and extracting global association of the features to finally generate a density map. According to the super-pixel network adopted by the invention, the pixels are groupedby utilizing the similarity of the characteristics among the pixels, and a small number of super-pixels are used for replacing a large number of pixels to represent the characteristics of the picture,so that the complexity is greatly reduced. The context-aware embedded network adopted by the method can extract global association between any two point features on the feature map, so the feature expression capability is enhan