Abstract:Visual defect detection, which is pivotal in industrial quality control, often requires extensive datasets for training deep-learning models. However, in industrial environments, the presence of multiple production batches, small lot sizes, and rapidly evolving task requirements make it challenging to acquire sufficient and diverse defect data. To address these challenges, this study introduces an innovative approach to data augmentation and defect detection under few-data conditions. Our strategy employs a two-stage feature-enhancement method complemented by an expert knowledge-based image-generation method. The former ensures robustness in color and richness in feature representation by focusing on local defect irregularities. The latter leverages both the core mechanistic understanding of experts and nonquantitative empirical insights, combined with a variable background domain, to generate rich images with diverse defect features. We evaluated the feasibility and effectiveness of the proposed method using a proprietary dataset curated by an automotive component manufacturing enterprise. The experimental results revealed that, across the five defect categories, training on 2 images achieved an F1-Score of 41.06%, which improved to 67.11% when the training dataset was expanded to 15 images. A comparative analysis with prevailing feature enhancement and image generation methods revealed our solution's superiority, bridging methodological drawbacks, and markedly eclipsing others in terms of enhancement potency. Furthermore, when benchmarked against additional public industrial datasets, the model exhibited exceptional adaptability and superior performance, thus highlighting its potential as a robust tool for quality control in manufacturing scenarios.

Synthetic Defect Generation for Display Front-of-Screen Quality Inspection: A Survey

A Novel Defect Detection System for Complex Freeform Surface Structures

Synthetic training data generation for deep learning based quality inspection

Surface Defect Detection of Mobile Phone Screen Based on Generative Adversarial Networks

Mask2Defect: A Prior Knowledge Based Data Augmentation Method for Metal Surface Defect Inspection

An Overview of Image Generation of Industrial Surface Defects

Particleboard Surface Defect Inspection Based on Data Augmentation and Attention Mechanisms

Coarse-to-Fine Few-Shot Defect Recognition with Dynamic Weighting and Joint Metric

Few-shot defect detection using feature enhancement and image generation for manufacturing quality inspection

Saliency-Based Defect Detection in Industrial Images by Using Phase Spectrum

A survey of real-time surface defect inspection methods based on deep learning

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

A Deep Learning-Based Surface Defect Inspection System Using Multiscale and Channel-Compressed Features

A Learning-Based Approach for Surface Defect Detection Using Small Image Datasets

Recent advances in surface defect inspection of industrial products using deep learning techniques

An Unsupervised-Learning-Based Approach for Automated Defect Inspection on Textured Surfaces

A Survey of Surface Defect Detection of Industrial Products Based on A Small Number of Labeled Data

Defect-GAN: High-Fidelity Defect Synthesis for Automated Defect Inspection

Surface Defect Detection Based on Deep Learning and Collaborative Cloud-edge Computation

DG-GAN: A High Quality Defect Image Generation Method for Defect Detection

Multi-stage Few-Shot Micro-Defect Detection of Patterned OLED Panel Using Defect Inpainting and Multi-Scale Siamese Neural Network