Crack-Detection-and-Segmentation-Dataset-for-UAV-Inspections

Dataset Overview

• Name: Crack‑Detection‑and‑Segmentation‑Dataset‑for‑UAV‑Inspections
• Purpose: Benchmark dataset for crack detection and segmentation, supporting multiple infrastructure crack types including road, bridge, and building cracks.
• Scale: Contains 11,298 crack images with fine pixel‑level annotations.
• Application: Can be used for unsupervised domain‑adaptive crack segmentation research.

Dataset Content

• Crack Types: Road, bridge, building cracks.
• Number of Images: 11,298.
• Annotations: Pixel‑level fine annotations.

Dataset Download

• Link: Google Drive
• Pre‑processing: Supports cropping, resizing, rotation, normalization, and flipping.

Dataset Applications

• Research Area: Unsupervised domain‑adaptive crack recognition.
• Model Transfer: Trained models can be transferred to unlabeled test sets with good performance.

Citation

@inproceedings{liu2019deep, title={Deep Learning Based Automatic Crack Detection and Segmentation for Unmanned Aerial Vehicle Inspections}, author={Liu, Kangcheng and Han, Xiaodong and Chen, Ben M}, booktitle={2019 IEEE International Conference on Robotics and Biomimetics (ROBIO)}, number={https://ieeexplore.ieee.org/document/896}, pages={381--387}, year={2019}, organization={IEEE} }

@article{liu2022industrial, title={Industrial uav‑based unsupervised domain adaptive crack recognitions: From system setups to real‑site infrastructural inspections}, author={Liu, Kangcheng and Chen, Ben M}, journal={IEEE Transactions on Industrial Electronics}, year={2022}, publisher={IEEE} }