Vision Mamba-based autonomous crack segmentation on concrete, asphalt, and masonry surfaces

TL;DR

This paper introduces a Vision Mamba-based framework for crack segmentation that achieves high accuracy and generalisation on various surfaces with significantly reduced computational complexity compared to CNNs and Transformers.

Contribution

The paper presents a novel VMamba-based architecture that improves crack segmentation accuracy and efficiency, addressing limitations of CNNs and Transformers.

Findings

Up to 74.5% fewer parameters compared to CNN models

Achieves up to 2.8% higher mDS than CNNs

Processes high-resolution images with 90.6% fewer FLOPs

Abstract

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.