Steel defect classification with Max-Pooling Convolutional Neural Networks

By Jonathan Masci, Ueli Meier, Dan Claudiu Ciresan, J. Urgen Schmidhuber, Gabriel Fricout. IEEE International Joint Conference on Neural Network, 2012

296 citations recorded by the source corpus.

We present a Max-Pooling Convolutional Neural Network approach for supervised steel defect classification. On a classification task with 7 defects, collected from a real production line, an error rate of 7% is obtained. Compared to SVM classifiers trained on commonly used feature descriptors our best net performs at least two times better. Not only we do obtain much better results, but the proposed method also works directly on raw pixel intensities of detected and segmented steel defects, avoiding further time consuming and hard to optimize ad-hoc preprocessing.

Subfields: Computer Vision and Pattern Recognition, Industrial and Manufacturing Engineering, Mechanical Engineering

View Paper: https://www.semanticscholar.org/paper/31534751dbd254fb7bf6ba2f43db5dfb6b54c6a7

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Industrial and Manufacturing Engineering

In Mechanical Engineering

Citations in the corpus

Segmentation-based deep-learning approach for surface-defect detection (2019, unknown): “The work by Masci et al (2012) showed that for surface-defect classiﬁcation the deep-learning approach can outperform classic machine-vi-sion approaches where hand-engineered features are combined with support vector machines.”
Deep convolutional neural networks for mobile capture device-based crop disease classification in the wild (2019, background): “…in dermatological (Esteva et al. (2017)) or histopathological images (Litjens et al. (2016); Arajo et al. (2017)), among others, to autonomous driving of cars ((Badrinarayanan et al. (2015) and drones (Gandhi et al. (2017))), or quality control in manufacturing industries (Masci et al. (2012)).”
Automatic Classification of Defective Photovoltaic Module Cells in Electroluminescence Images (2018, background): “[11] who proposed an end-to-end max-pooling CNN for classifying steel defects.”
Review of vision-based steel surface inspection systems (2014, background): “Steel surfaces exhibit many types of defects, and NN-BP has been extensively used to classify such defects [31,36,54,59,63,72,78,80].”
Deep convolutional neural networks for detection of rail surface defects (2016, methodology): “Moreover, in [19], a max-pooling convolutional neural network is applied for steel defect classification.”
Three-dimensional convolutional neural network (3D-CNN) for heterogeneous material homogenization (2020, unknown): “Some of the recent applications of DL approaches on material science include material classification [35, 36], defect classification [37–39], microstructure identification [40, 41], microstructure reconstruction [42, 43], composite strength prediction [44], etc.”
A review of machine learning for the optimization of production processes (2019, methodology): “Various meth-ods were developed for the production of semiconductors and screen glasses [40, 43, 105], ceramics and tiles [47], and miscellaneous processes relating to metal parts [42, 63, 88, 113, 121].”
Advanced Steel Microstructural Classification by Deep Learning Methods (2017, methodology): “10 used CNNs to find defects in steel.”
A fast learning algorithm for image segmentation with max-pooling convolutional networks (2013, background): “MPCNN scale well to large images and excel in many object recognition [9, 8, 10, 7] and segmentation [3, 6, 5] benchmarks.”
Design of deep convolutional neural network architectures for automated feature extraction in industrial inspection (2016, background): “The presented network design heuristics are strongly inspired by [14,15] due to its simplicity and effectiveness.”
Surface defect saliency of magnetic tile (2018, background): “At the same time deep neural networks have become dominant on the surface defect detection, achieving stateof-the-art performance on classification tasks [15], [16].”
State of the Art in Defect Detection Based on Machine Vision (2021, unknown): “[178] presented a max-pooling CNN approach.”