CTCD Complex Tunnel Crack Dataset: 292 Highway Tunnel Lining Images with Pixel-Level Crack Masks (CraSAM)
- Provider
- Shiwei Luo, Xiongyao Xie, Biao Zhou, Kun Zeng and Jun Guo
- License
- CC-BY-4.0 (source)
- Size
- 586 files
- Stored at
- Hugging Face (https://huggingface.co/datasets/shiweiluo99/tunnel-crack-segmentation-dataset)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Computer Vision
- Source
- https://doi.org/10.1155/stc/6269747
About this dataset
The Complex Tunnel Crack Dataset (CTCD) is a civil-infrastructure inspection set of 292 tunnel lining images (256x256 pixels) with pixel-level crack segmentation masks, collected from 52 highway tunnels in Yunnan and Zhejiang, China, with a high-speed tunnel inspection vehicle. Crack pixels were annotated by hand in Photoshop with a 1 to 2 pixel brush sized to the crack width. It was introduced in the CraSAM paper, which builds a crack detection network on the Segment Anything visual foundation model, to test how crack segmentation generalizes across scenarios. The repository holds 584 BMP files (images and masks) in one folder, about 80 MB (measured 2026-11-21); load with Hugging Face datasets or PIL. Use it for crack segmentation, tunnel defect inspection and evaluating segmentation models on unseen infrastructure scenes. Images and annotations are by the authors and released under CC-BY-4.0; cite Luo et al., Multiscenario Generalization Crack Detection Network Based on the Visual Foundation Model, Structural Control and Health Monitoring 2025, doi:10.1155/stc/6269747.
How to load this dataset in Python
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="shiweiluo99/tunnel-crack-segmentation-dataset", filename="Complex Tunnel Crack Dataset (CTCD)/train/1001-1.bmp", repo_type="dataset")
print(path)Data structure
Files
Example file: Complex Tunnel Crack Dataset (CTCD)/train/1001-1.bmp
Analyze this dataset without downloading it
On Scigantic, CTCD Complex Tunnel Crack Dataset: 292 Highway Tunnel Lining Images with Pixel-Level Crack Masks (CraSAM) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Computer Vision datasets
- MMLA Pose: Eight-Way Orientation Labels for Zebras in Drone Footage
- Vipe Dynpose 100 Kpp
- Llama Nemotron VLM Dataset V 1
- IntPhys 2
- TartanAir V2: multi-sensor synthetic dataset for SLAM, navigation and robot perception
- Nemotron Research Lgt
- Vipe Wild Sdg 1 M
- GeckoNum: Numerical Reasoning Benchmark for Text-to-Image Models, Generated Images from Imagen, Muse and DALL-E 3 with Counting Prompts (Google DeepMind)