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).

Automatic Classification of Defective Solar Panels in Electroluminescence Images Based on Random Connection Network

Deep learning based automatic defect identification of photovoltaic module using electroluminescence images

Deep-Learning-Based Automatic Detection of Photovoltaic Cell Defects in Electroluminescence Images

Automatic Classification of Defective Photovoltaic Module Cells in Electroluminescence Images

CNN based automatic detection of photovoltaic cell defects in electroluminescence images

Solar Cells Surface Defects Detection Based on Deep Learning

Defect Detection with Generative Adversarial Networks for Electroluminescence Images of Solar Cells

Accurate Detection and Intelligent Classification of Solar Cells Defects based on Photoluminescence Images : A Novel Study on the Optimized YOLOv5 Model

Deep Learning Based Model for Defect Detection of Mono-Crystalline-Si Solar PV Module Cells in Electroluminescence Images Using Data Augmentation

Automatic defect identification of PV panels with IR images through unmanned aircraft

Detection of solar panel defects based on separable convolution and convolutional block attention module

Classification of solar panel technology and photovoltaic cell status applying machine learning to electroluminescence images

Solar photovoltaic module defect detection based on deep learning

A Deep Two-Stage Scheme for Polycrystalline Micro-Crack Detection

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

CNN-based Model for Cell Extraction from PV Modules with EL Images for PV Defects Detection

A Definition Rule for Defect Classification and Grading of Solar Cells Photoluminescence Feature Images and Estimation of CNN-Based Automatic Defect Detection Method

Defect recognition of photovoltaic panels based on adaptive dimensional feature aggregation convolutional neural network

A lightweight network for photovoltaic cell defect detection in electroluminescence images based on neural architecture search and knowledge distillation

Defect object detection algorithm for electroluminescence image defects of photovoltaic modules based on deep learning

Deep Learning-Based Algorithm for Multi-Type Defects Detection in Solar Cells with Aerial EL Images for Photovoltaic Plants