To achieve this, we propose an RGB-T crowd counting network based on global-local multimodal feature fusion (GLFNet). Specifically, we first use a multihead attention mechanism to fuse global ...