PVC pipe size detection method and device based on high-resolution semantic segmentation

The invention discloses a PVC pipe size detection method and device based on high-resolution semantic segmentation, and the method comprises the steps: S01, obtaining an original image, and carrying out the transformation of the original image, and obtaining a to-be-segmented image; s02, the image to be segmented is input into a high-resolution semantic segmentation model for semantic segmentation, the high-resolution semantic segmentation model is based on an ISDNet model, in an RAF module, attention layer parameters are generated for a shallow feature map and a deep feature map through two parallel branches respectively, attention layer parameters are fused through MLP layer fusion, and the attention layer parameters of shallow channels and deep channels are fused; fusing with a space attention module to obtain a final space attention parameter; step S03, extracting a pipeline edge region from a high-resolution semantic segmentation result; and step S04, extracting size information of the PVC pipeline from