The front-end of the PDDNet chooses the first 13 layers of GhostNet to extract the crowd feature, and the back-end of the PDDNet is implemented with the proposed lightweight pyramidal convolution modules (LPC) to extract the multi-scale features. Finally, the extracted multi-scale features ...
By combining depth-wise separable convolution residual unit and self-attention, the authors propose a deep separable residual asymmetric self-attention network for road extraction tasks, which can obtain more complete road information in complex scenes; Lightweight asymmetric self-attention effectively reduc...