Segment Any Crack: Deep Semantic Segmentation Adaptation for Crack Detection

TL;DR

The paper tackles domain shift in crack segmentation by introducing a parameter-efficient selective fine-tuning approach that tunes only normalization parameters, applied to the Segment Anything Model (SAM) and other baselines to form the Segment Any Crack (SAC). Across OmniCrack30k and three zero-shot datasets, SAC achieves a F1-score of $61.22\%$ and an IoU of $44.13\%$, outperforming full fine-tuning and several PEFT baselines while using only about $0.0458\%$ of the total parameters. The results demonstrate strong generalization to unseen environments and substantial computational efficiency, highlighting normalization-focused adaptation as a practical strategy for infrastructure health monitoring with vision foundation models. This work suggests broader applicability of layer-normalization tuning for domain adaptation in large segmentation models and sets a benchmark for efficient crack detection in real-world settings.

Abstract

Image-based crack detection algorithms are increasingly in demand in infrastructure monitoring, as early detection of cracks is of paramount importance for timely maintenance planning. While deep learning has significantly advanced crack detection algorithms, existing models often require extensive labeled datasets and high computational costs for fine-tuning, limiting their adaptability across diverse conditions. This study introduces an efficient selective fine-tuning strategy, focusing on tuning normalization components, to enhance the adaptability of segmentation models for crack detection. The proposed method is applied to the Segment Anything Model (SAM) and five well-established segmentation models. Experimental results demonstrate that selective fine-tuning of only normalization parameters outperforms full fine-tuning and other common fine-tuning techniques in both performance and computational efficiency, while improving generalization. The proposed approach yields a SAM-based model, Segment Any Crack (SAC), achieving a 61.22\% F1-score and 44.13\% IoU on the OmniCrack30k benchmark dataset, along with the highest performance across three zero-shot datasets and the lowest standard deviation. The results highlight the effectiveness of the adaptation approach in improving segmentation accuracy while significantly reducing computational overhead.

Figures (11)

• Figure 1: The proposed layer normalization fine-tuning method illustrated on the Segment Anything Model (SAM). Trainable modules are colored in red and frozen modules in grey.
• Figure 2: Samples of the OmniCrack30k dataset. Note that Khanh. and Rissb. denote Khanh11k and Rissbilder, respectively.
• Figure 3: Samples from the datasets used for zero-shot evaluations. The first, second, and third rows, correspond to the Road420, Facade390, and Concrete3k datasets, respectively.
• Figure 4: Ablation results of adapting the Segment Anything Model (SAM) using different fine-tuning methods. The proposed normalization fine-tuning method achieves the highest performance among other methods with a relatively lower number of trainable parameters.
• Figure 5: Comparison of the untuned and fine-tuned Segment Anything Model (SAM) segmentation results on three sample input images. In each case, the first row illustrates the prediction mask of the unturned SAM and the second row demonstrates the prediction output of the fine-tuned model.