Reinforced concrete structure segmentation dataset 2119

This dataset is a comprehensive, pixel-level annotated image dataset for multi-type damage detection and segmentation in reinforced concrete (RC) structures.

This dataset is freely available for academic and research purposes, including direct use in training, validation, benchmarking, and comparative studies of damage detection and segmentation algorithms.

If you use this dataset in any publication, report, or derivative work, please cite the following related papers:

Wang, J., & Ueda, T. (2025). Automatic damage detection and segmentation using deep learning algorithms in reinforced concrete structure inspections. Structural Concrete, 26(5), 5511–5534.

Wang, J., Wang, Z., Wang, Y., & Li, Z. (2025). Automated multi-type damage detection framework in reinforced concrete structures via data augmentation and deep segmentation networks. Journal of Civil Structural Health Monitoring, 15(8), 3861–3884.

A total of 2,119 reinforced concrete (RC) structure damage images were selected from public datasets and manually annotated using distinct color labels: gray for background, red for concrete cracks, green for concrete spalling, magenta for rebar exposure, blue for rebar corrosion, and yellow for concrete crushing. As illustrated in Fig. 1f, multiple damage types may coexist within a single image; consequently, the total number of annotated damage instances far exceeds the number of images. After annotation, the number of instances in each damage category is 1,883 for concrete cracks, 282 for concrete spalling, 89 for rebar exposure, 832 for rebar corrosion, and 88 for concrete crushing.