Adaptive multi-scale context aggregation method for crowd counting

The invention provides a self-adaptive multi-scale context aggregation method for crowd counting, and the method comprises the steps: inputting a sample picture into a backbone network, and extracting a feature map of which the size is j times of the resolution of an input image; inputting the extracted feature maps into a plurality of multi-scale context aggregation modules in a cascade form, and extracting and adaptively aggregating multi-scale context information to obtain multi-scale context features; performing convolutional layer processing on the generated multi-scale context features to generate a density map; and carrying out integral summation on the density map to obtain a predicted number of people. According to the method, multi-scale information is effectively extracted, the problem of non-uniform head size is solved, useful context information is adaptively selected and aggregated through a channel attention mechanism, information redundancy is avoided, more accurate density estimation can be a