[2603.01361] MixerCSeg: An Efficient Mixer Architecture for Crack Segmentation via Decoupled Mamba Attention

Computer Science > Computer Vision and Pattern Recognition arXiv:2603.01361 (cs) [Submitted on 2 Mar 2026] Title:MixerCSeg: An Efficient Mixer Architecture for Crack Segmentation via Decoupled Mamba Attention Authors:Zilong Zhao, Zhengming Ding, Pei Niu, Wenhao Sun, Feng Guo View a PDF of the paper titled MixerCSeg: An Efficient Mixer Architecture for Crack Segmentation via Decoupled Mamba Attention, by Zilong Zhao and Zhengming Ding and Pei Niu and Wenhao Sun and Feng Guo View PDF HTML (experimental) Abstract:Feature encoders play a key role in pixel-level crack segmentation by shaping the representation of fine textures and thin structures. Existing CNN-, Transformer-, and Mamba-based models each capture only part of the required spatial or structural information, leaving clear gaps in modeling complex crack patterns. To address this, we present MixerCSeg, a mixer architecture designed like a coordinated team of specialists, where CNN-like pathways focus on local textures, Transformer-style paths capture global dependencies, and Mamba-inspired flows model sequential context within a single encoder. At the core of MixerCSeg is the TransMixer, which explores Mamba's latent attention behavior while establishing dedicated pathways that naturally express both locality and global awareness. To further enhance structural fidelity, we introduce a spatial block processing strategy and a Direction-guided Edge Gated Convolution (DEGConv) that strengthens edge sensitivity under irre...