Aiming to quickly, accurately, and automatically recognize various defects (leakage, crack, spalling) from massive metro tunnel lining image data, a two-step deep learning-based framework is proposed.
In this repository, the datasets for metro tunnel lining image classification and multi-defect detection are shared for further research. Meanwhile, the algorithms proposed in the paper and related weight files are also shared.
A metro tunnel lining image classification dataset
Image classification algorithms and model weights files generated from classification experiments
A metro tunnel lining multi-defect dataset for object detection
Object detection algorithms and model weights files generated from object detection experiments
This dataset consists of defect-free images and defective images of metro tunnel lining, which can be download from https://pan.baidu.com/s/1wD-XEHMQ8uXxJMpPNfKAKQ. Password is 3lyt.
The source code for TLCNet is placed in the folder named TLCNet-main. The relevant model weight files can be downloaded from https://pan.baidu.com/s/1Uniz0kqg8PuFW3Ygv3I4rQ. Password: 6ui4.
This dataset consists of original metro tunnel lining defective images and accurate labels, which can be download from https://pan.baidu.com/s/1ph7havsMRFy3EABhKiSV4Q. Password is 728j.
The source code for TDDNet is placed in the folder named TDDNet-main. The relevant model weight files can be downloaded from https://pan.baidu.com/s/1hixgINiUfARwUx1xvKENow. Password: 8t15.
The model generalization experiment in this research was conducted based on some open-source tunnel lining images from three studies[1-3]. [1]https://doi.org/10.1109/ICIP40778.2020.9190900 [2]https://doi.org/10.1016/j.tust.2023.105428 [3]https://doi.org/10.1016/j.tust.2020.103524 Based on the original images, our research team has made labels in VOC format to evaluate the performance of the classification model and the target detection model. The dataset can be downloaded from the link https://pan.baidu.com/s/1y5v4Ihpn9LFNQiSkLfdAKw. Password is 1h20.
Yong Feng, Shi-Jin Feng, Xiao-Lei Zhang, Dong-Mei Zhang, Yong Zhao
A two-step deep learning-based framework for metro tunnel lining defect recognition
Tunnelling and Underground Space Technology
Volume 150
2024
105832
ISSN 0886-7798
https://doi.org/10.1016/j.tust.2024.105832.
https://www.sciencedirect.com/science/article/pii/S0886779824002505