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

Researchers have developed MixerCSeg, a new AI architecture for crack segmentation that combines CNN, Transformer, and Mamba-based approaches to achieve state-of-the-art performance with high efficiency. The model uses only 2.05 GFLOPs and 2.54M parameters while outperforming existing methods on crack detection benchmarks.

• →MixerCSeg introduces a hybrid architecture combining CNN, Transformer, and Mamba approaches for superior crack segmentation.
• →The model achieves state-of-the-art performance with minimal computational requirements (2.05 GFLOPs, 2.54M parameters).
• →TransMixer explores Mamba's attention behavior while creating pathways for both local and global feature awareness.
• →Direction-guided Edge Gated Convolution (DEGConv) enhances edge sensitivity for irregular crack geometries.
• →The architecture addresses gaps in existing models by capturing spatial, structural, and sequential information simultaneously.