Convolutional networks for voting-based anomaly classification in metal surface inspection

Abstract

Automated Visual Inspection (AVI) systems for metal surface inspection is increasingly used in industries to aid human visual inspectors for classification of possible anomalies. For classification, the challenge lies in having a small and specific dataset that may easily result in over-fitting. As a solution, we propose to use deep Convolutional Neural Networks (ConvNets) learnt from the large ImageNet dataset [9] for image representations via transfer learning. Since a small dataset cannot be used to fine-tune a ConvNet due to overfitting, we also propose a Majority Voting Mechanism (MVM), which fuses the features extracted from the last three layers of ConvNets using Support Vector Machine (SVM) classifiers. This classification framework is effective where no prior knowledge of the best performing ConvNet layers is needed. This also allows flexibility in the choice of ConvNet used for feature extraction. The proposed method not only outperforms state-of-the-art traditional hand-crafted features in terms of classification but also obtains good results compared to other deep features with selected best layers on several anomaly and texture datasets.

PGA-Net: Pyramid Feature Fusion and Global Context Attention Network for Automated Surface Defect Detection

“…Image-level defect classification: Masci et al [30] proposed a multi-scale pyramidal pooling network for classification of steel defect, which didn't require the size of all images to be equal. Natarajan et al [31] proposed a flexible multi-layered deep feature extraction through transfer learning and SVM classifiers, which overcome the problem of over-fitting caused by small datasets. He et al [32] proposed a semi-supervised model of CNN for feature extraction and fed the representation features into a classifier for classification of steel surface defect.…”

Automated Visual Defect Classification for Flat Steel Surface: A Survey

“…However, the performance of CNN-based methods mainly depend on plentiful training samples, which stunts the utilization of CNNs in industrial scenes with small datasets. At present, transfer learning makes full use of the previously labelled data and guarantees the precision of the model on new tasks with limited training samples [107], which broadens the application prospect of the CNN. CNNs are the core of deep learning methods, and many more features and applications are worth exploring and should be given more attention.…”

Defect Detection and Recognition of Mobile Phone Membrane Based on Convolutional Neural Network

“…In the past decade, many researches have been devoted to the application of machine vision technology in surface defect detection, and The neural network algorithm is classifying the target, classification of natural scenes got good results [21]. A flexible multi-layer deep feature extraction framework based on CNN is proposed to detect anomalies in anomalous data sets [22].Lin et al Established a convolutional neural network (CNN) for fault checking of LED chips [23]. Defective areas are located using activation-like mapping technology without the need for annotation at the human area level.…”