Abstract:Visual surface defect inspection provides an important tool for product quality assessment in a wide range of industrial applications. In recent years, diverse convolutional neural network (CNN) models have been devel-oped for the high-accuracy image classification of natural objects. However, it remains a challenging task to develop deep-learning-based approaches for surface defect inspection of industrial products. The main challenge is that surface defects of industrial products occur in low probability, therefore it is impractical to build a large image dataset containing different types of defects to train/fine-tune deep CNN models to achieve satisfactory performance when applied to industrial inspection tasks. To overcome the above-mentioned problem, we build a light-weight defect classification model based on the pre-trained SqueezeNet architecture and present three effective techniques to achieve high-accuracy defect classification using very few collected defective training samples. First, we experimentally demonstrate that it is reasonable to freeze shallower convolutional layers and only fine-tune deeper layers in the last convolutional stage when there are not enough training samples. Second, we integrate batch-size independent Group Normalization (GN) into SqueezeNet to compute stable statistics based on a limited number of defective samples for training the classification model. Third, we add an auxiliary task to perform the similarity comparison of two defective samples during the training phase, significantly increasing the size of the training data and generating more distinctive defect-specific representations. The experimental results on the NEU-CLS dataset and USB-FS dataset verify the effectiveness of the proposed method. Our method can obtain accurate defect classification results when very few defective training samples are provided, i.e., about 97.69% and 82.92% average classification accuracy on the NEU-CLS and USB-FS datasets, respectively, under the 5-shot setting (5 images per defect category).

Self-supervised assisted multi-task learning network for one-shot defect segmentation with fake defect generation

A Sub-Region Unet for Weak Defects Segmentation with Global Information and Mask-Aware Loss.

Efficient surface defect detection using self-supervised learning strategy and segmentation network

An Adaptive Image Segmentation Network for Surface Defect Detection

AnomalySeg: Deep Learning-Based Fast Anomaly Segmentation Approach for Surface Defect Detection

Multi-modal background-aware for defect semantic segmentation with limited data

Few-shot Defect Recognition of Metal Surfaces Via Attention-Embedding and Self-Supervised Learning

LSA-Net: Location and shape attention network for automatic surface defect segmentation

A CAM-based Weakly Supervised Method for Surface Defect Inspection

A nondestructive automatic defect detection method with pixelwise segmentation

A robust weakly supervised learning of deep Conv-Nets for surface defect inspection

A Pixel-Level Segmentation Convolutional Neural Network Based on Global and Local Feature Fusion for Surface Defect Detection

Surface Defect Detection and Semantic Segmentation with a Novel Lightweight Deep Neural Network

Interactive Defect Segmentation in X-Ray Images Based on Deep Learning

Detail-semantic guide network based on spatial attention for surface defect detection with fewer samples

An Efficient End-to-End Multitask Network Architecture for Defect Inspection

Few-Shot Defect Segmentation Leveraging Abundant Normal Training Samples Through Normal Background Regularization and Crop-and-Paste Operation

Semi-supervised auxiliary learning for surface defect detection and segmentation of injection-molded products from small image datasets

Multi-position industrial defect inspection using self-training siamese networks with mix strategies

An Effective Industrial Defect Classification Method under the Few-Shot Setting Via Two-Stream Training

Automated Detection of Defects with Low Semantic Information in X-ray Images Based on Deep Learning