An enhanced network for extracting tunnel lining defects using transformer encoder and aggregate decoder
Abstract: Highlights•A novel enhanced deep learning architecture combining Transformer with convolution is proposed.•An advanced Transformer backbone capable of exploring both local fine-grained and global contextual information.•A well-designed decoder enhances the perceptual range to fully extract defect regions with multiple granularities.
External IDs:dblp:journals/aeog/GuoHLAWHC25
Loading