huangyebiaoke / steel-pipe-weld-defect-detection

Deep Learning Based Steel Pipe Weld Defect Detection

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Steel Pipe Weld Defect Detection

This repository contains the codes & dataset for the paper: Dingming Yang, Yanrong Cui, Zeyu Yu & Hongqiang Yuan. (2021). Deep Learning Based Steel Pipe Weld Defect Detection. [paper] [arxiv] [code]

result

Run Locally

Clone the project

git clone https://github.com/huangyebiaoke/steel-pipe-weld-defect-detection

Go to the project directory

cd steel-pipe-weld-defect-detection

Install dependencies

pip install -r requirements.txt

Download dataset from Releases and unzip the file to the current directory

wget https://github.com/huangyebiaoke/steel-pipe-weld-defect-detection/releases/download/1.0/steel-tube-dataset-all.zip
unzip steel-tube-dataset-all.zip

Start training model

py ./yolov5/train.py

Dataset

You can get the dataset from Releases which with YOLO and PASCAL VOC 2007 Format in the zip file.

Sample distribution

sample-distribution

EN air-hole bite-edge broken-arc crack hollow-bead overlap slag-inclusion unfused
ZH 气孔 咬边 断弧 裂缝 夹珠 焊瘤 夹渣 未融合
Label 0 1 2 3 4 5 6 7
Number 5191 35 458 119 229 223 120 408

Dataset preview

samples-data-show

Dataset analysis

sample-data-analysis-v3-en

Citation

If you use the code or dataset provided in this repository, please cite this work as follows:

@article{doi:10.1080/08839514.2021.1975391,
author = {Dingming Yang and Yanrong Cui and Zeyu Yu and Hongqiang Yuan},
title = {Deep Learning Based Steel Pipe Weld Defect Detection},
journal = {Applied Artificial Intelligence},
volume = {0},
number = {0},
pages = {1-13},
year  = {2021},
publisher = {Taylor & Francis},
doi = {10.1080/08839514.2021.1975391},
URL = {https://doi.org/10.1080/08839514.2021.1975391},
eprint = {https://doi.org/10.1080/08839514.2021.1975391}
}

Related works

Acknowledgements

License

GPL-3.0

About

Deep Learning Based Steel Pipe Weld Defect Detection

License:GNU General Public License v3.0


Languages

Language:Jupyter Notebook 95.7%Language:Python 4.2%Language:Dockerfile 0.0%Language:TeX 0.0%Language:Shell 0.0%