BGCrack

This repository is the official PyTorch implementation of the Boundary Guidance Crack Segmentation Model (BGCrack).

🍇 Paper

• Journal Version | arXiv Version
• Title: Crack Segmentation on Steel Structures Using Boundary Guidance Model

🍎 Dataset:

• The Steelcrack dataset is available at Civil-dataset.

🛴 Updates

• 2026/05/09: Apologies for the delayed release of the code. It is now available!
• 2024/03/06: 🥂🥂 The paper is now available online! Link → Journal Paper.
• 2024/02/27: 🥳🥳 Our paper, Crack Segmentation on Steel Structures Using Boundary Guidance Mode, has been accepted by Automation in Construction after an 8-month review process!
• 2023/06/15: The preprint of our paper has been submitted to arXiv. Link → arXiv Paper.
• 2023/05/10: CSNSS has been renamed to BGCrack.
• 2022/10/17: This repository is now live! Its previous name was CSNSS (Crack Segmentation Network for Steel Structures).

🚀 Getting Started

1. Requirements

Recommended versions are
    * python = 3.8.15
    * pytorch = 1.12.1
    * CUDA 11.6.2 and CUDNN 8.6.0  
Other requirements can be found in requirements.txt.

2. Installation

git clone https://github.com/hzlbbfrog/BGCrack
cd BGCrack
pip install -r requirements.txt

Or, you can directly "Download ZIP".

3. Prepare the dataset

Download the dataset from Civil-dataset and organize the folder structure as follows:

BGCrack/
├── Dataset/
│   ├── Steelcrack/
│   │   ├── Train/
│   │   │   ├── images/
│   │   │   ├── masks/
│   │   │   └── edges/
│   │   ├── Validation/
│   │   │   ├── images/
│   │   │   ├── masks/
│   │   │   └── edges/
│   │   ├── Test/
│   │   │   ├── images/
│   │   │   ├── masks/
│   │   │   └── edges/

4. Training and validation

To train the BGCrack model, run the following command:

python train_BGCrack_2024.py --dataset=steel_cracks_with_edge --modelname=BGCrack --batchsize=9 --epoch=70 --lr=0.006

Logs and model checkpoints will be created in ./Result_log and ./Checkpoints/ respectively.

5. Test

To evaluate the model, run the testing script (replace <YOUR_TEST_EPOCH> with the specific epoch you want to evaluate):

python test_BGCrack_2024.py --dataset=steel_cracks_with_edge --modelname=BGCrack --test_epoch=<YOUR_TEST_EPOCH>

🎯 Method

BGCrack (Boundary Guidance Crack Segmentation Model) is designed for crack segmentation on steel structures. It features modules for boundary guidance to explicitly incorporate crack edge information, leading to precise crack delineation. The model integrates deep feature extraction mechanisms like MobileViT and attention gates.

🎖️ Results on Steelcrack dataset

Method	mi IoU (%)	mi Dice (%)	#Param. (M)	MACs (G)
U-Net	68.49	75.13	7.77	55.01
U-Net++	72.23	78.37	9.16	138.63
Attention U-Net	71.25	77.54	34.88	266.54
CE-Net	76.00	81.54	29.00	35.60
DeepLabv3+ (MobileNetv2)	68.22	71.07	5.81	29.13
DeepLabv3+ (Xception)	67.40	71.48	54.70	83.14
DeepLabv3+ (ResNet-101)	69.04	69.45	59.34	88.84
SCRN	73.23	78.91	25.23	31.92
TransUNet	64.34	72.55	67.87	129.96
CrackSeU-B	70.42	80.50	3.19	11.22
CrackSeU-L	71.66	81.24	4.62	28.22
DconnNet	74.73	83.40	28.38	24.79
BGCrack V1	77.16	85.33	2.32	15.76

🥰 Cite BGCrack!

If you have any problems, please do not hesitate to contact us! You are very welcome to cite our paper!
The BibTeX entry of the paper is as follows:

@article{BGCrack,
title = {Crack segmentation on steel structures using boundary guidance model},
journal = {Automation in Construction},
volume = {162},
pages = {105354},
year = {2024},
issn = {0926-5805},
doi = {https://doi.org/10.1016/j.autcon.2024.105354},
url = {https://www.sciencedirect.com/science/article/pii/S0926580524000906},
author = {Zhili He and Wang Chen and Jian Zhang and Yu-Hsing Wang},
keywords = {Crack inspection, Deep learning, Boundary guidance method, Benchmark dataset}
}

🌹 Acknowledgements

This repo benefits from SCRN, FcaNet, and MobileViT. Thanks for their wonderful works!