Research on image segmentation method based on depth-separable double-void pyramid pooling networks

In recent years, image segmentation technology has been widely used in remote sensing image analysis, medical image processing and target detection. However, the traditional image segmentation methods have been exposed to the problems of high computational complexity and low sensitivity to details; the deep learning methods have the advantage of automatic feature extraction but have the disadvantage of high consumption of computational resources. To address the above problems, this study innovatively proposes an image segmentation method based on depth-separable bi-null pyramid pooling network, aiming to significantly improve the accuracy of image segmentation while effectively reducing the computational overhead. The research is carried out in three aspects, firstly, the introduction of depth-separable convolution reduces the computational overhead and improves the computational efficiency. Secondly, dual null pyramid pooling is introduced to enhance the network's ability to perceive features at different scales. Finally, the decoder structure is optimized to improve segmentation accuracy by recovering the resolution through inverse convolution and jump connection.