adipandas / one-shot-steel-surfaces

One-Shot Recognition of Manufacturing Defects in Steel Surfaces

One-Shot Recognition of Manufacturing Defects in Steel Surfaces

This repository contains the codes for the paper:
Deshpande, A. M., Minai, A. A., & Kumar, M. (2020). One-Shot Recognition of Manufacturing Defects in Steel Surfaces. [arxiv] [paper] [website] [code]

Network Architecture

Training Loss curve

Requirements

numpy
scipy
matplotlib
torch
torchvision
scikit-learn
imutils
opencv-python
Pillow
jupyterlab

Do the following in given order to install all the packages

Create a python virtual environment, preferrablely using Anaconda.
Download anaconda and install from here.
To create the virtual environment, open terminal (anaconda prompt) and execute:

conda create -n steel_p36 python=3.6

Activate python environment in your terminal:

conda activate steel_p36

Run the following commands in your terminal to install all the dependencies.

pip install numpy scipy matplotlib
pip install jupyterlab
pip install Pillow
pip install opencv-contrib-python
pip install -U scikit-learn
pip install torch torchvision

Dataset

Dataset Reference: Song, K., Yan, Y. (2013). A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects. Applied Surface Science, 285, 858-864.

You can get the dataset from:

• NEU Steel Surface defect dataset: website

I have noticed that there are issues such as dataset website is not reachable. In that case you can get the dataset from the following Google Drive link:

• Google Drive

Citation

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

@article{deshpande20201064,
title={{One-Shot Recognition of Manufacturing Defects in Steel Surfaces}},
journal= {Procedia Manufacturing},
volume= {48},
pages= {1064 - 1071},
year= {2020},
note= {48th SME North American Manufacturing Research Conference, NAMRC 48},
issn= {2351-9789},
doi= {https://doi.org/10.1016/j.promfg.2020.05.146},
url= {http://www.sciencedirect.com/science/article/pii/S2351978920315985},
author= {Aditya M. Deshpande and Ali A. Minai and Manish Kumar},
keywords= {Computer Vision, Deep Learning, Metallic Surface, Convolutional Neural Network, Defect Detection, One-shot recognition, Industrial Internet of Things, Cyber-physical systems, Siamese neural network, Few-shot learning},
}

one-shot-steel-surfaces's Issues

GPU 训练

请问我将模型放到GPU训练后，为何代码没有反应？