基于Vision Mamba的混凝土、沥青和砖石表面自主裂缝分割
Vision Mamba-based autonomous crack segmentation on concrete, asphalt, and masonry surfaces
AI总结：
针对CNNs泛化性弱、Transformer计算复杂度高的问题，本文提出基于Vision Mamba的裂缝分割框架，在精度相当的前提下大幅降低参数量与浮点运算量。
摘要：
卷积神经网络（CNNs）和Transformer在特定条件下的裂缝检测任务中已展现出先进精度。然而，固定的局部注意力会削弱CNNs的泛化能力，而全局自注意力的二次复杂度限制了Transformer的实际部署。鉴于新一代架构Mamba的出现，本文提出一种基于Vision Mamba（VMamba）的框架，用于混凝土、沥青和砖石表面的裂缝分割，具备高精度、强泛化性且计算复杂度更低。与具有代表性的基于CNN的模型相比，集成了VMamba的编码器-解码器网络参数量减少15.6%-74.5%，同时平均戴斯系数（mDS）最高提升2.8%，性能与基于Transformer的模型相当。此外，该基于VMamba的编码器-解码器网络处理高分辨率图像输入时，浮点运算量最高可降低90.6%。
英文摘要：
Convolutional neural networks (CNNs) and Transformers have shown advanced accuracy in crack detection under certain conditions. Yet, the fixed local attention can compromise the generalisation of CNNs, and the quadratic complexity of the global self-attention restricts the practical deployment of Transformers. Given the emergence of the new-generation architecture of Mamba, this paper proposes a Vision Mamba (VMamba)-based framework for crack segmentation on concrete, asphalt, and masonry surfaces, with high accuracy, generalisation, and less computational complexity. Having 15.6% - 74.5% fewer parameters, the encoder-decoder network integrated with VMamba could obtain up to 2.8% higher mDS than representative CNN-based models while showing about the same performance as Transformer-based models. Moreover, the VMamba-based encoder-decoder network could process high-resolution image input with up to 90.6% lower floating-point operations.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
