Weld defect segmentation algorithm based on improved U-net

The traditional convolutional neural network has the problems of insufficient accuracy and insufficient use of context features in weld segmentation. Therefore, an improved U-Net based weld segmentation algorithm is proposed, which uses VGG network as the coding part to enhance the feature extraction capability. At the same time, CBAM (Convolutional Block Attention Module) and SENet (Squeeze and Excitation Networks) are introduced to enable the model to dynamically adjust the feature representation to highlight important channels and spatial locations. Thus improve the accuracy of classification and segmentation. The experimental results show that the accuracy and robustness of the improved model are significantly improved under complex background.