Sequential PatchCore: Anomaly Detection for Surface Inspection using Synthetic Impurities

2025-01-16

AI Summary

In industrial surface inspection, automatic visual inspection suffers from degraded performance due to subtle contaminants (e.g., water stains, fingerprints), while high-resolution image training is constrained by GPU memory and cross-dataset knowledge transfer remains challenging. To address these issues, this work proposes: (1) a novel procedural water-stain synthesis method explicitly modeling contaminants, generating photorealistic, high-fidelity defect samples; (2) Sequential PatchCore—a lightweight, sequential core-set construction framework for anomaly detection that alleviates GPU memory bottlenecks, enabling end-to-end training on hundred-megapixel images using consumer-grade hardware; and (3) a two-stage learning strategy combining synthetic-data pretraining with real-data fine-tuning, complemented by defect-level recall evaluation. Experiments demonstrate substantial improvements in robustness for detecting fine-grained defects and validate effective cross-version core-set transfer learning.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Large pretrained models with web data
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

The appearance of surface impurities (e.g., water stains, fingerprints, stickers) is an often-mentioned issue that causes degradation of automated visual inspection systems. At the same time, synthetic data generation techniques for visual surface inspection have focused primarily on generating perfect examples and defects, disregarding impurities. This study highlights the importance of considering impurities when generating synthetic data. We introduce a procedural method to include photorealistic water stains in synthetic data. The synthetic datasets are generated to correspond to real datasets and are further used to train an anomaly detection model and investigate the influence of water stains. The high-resolution images used for surface inspection lead to memory bottlenecks during anomaly detection training. To address this, we introduce Sequential PatchCore - a method to build coresets sequentially and make training on large images using consumer-grade hardware tractable. This allows us to perform transfer learning using coresets pre-trained on different dataset versions. Our results show the benefits of using synthetic data for pre-training an explicit coreset anomaly model and the extended performance benefits of finetuning the coreset using real data. We observed how the impurities and labelling ambiguity lower the model performance and have additionally reported the defect-wise recall to provide an industrially relevant perspective on model performance.

Problem

• Visual Detection Degradation
• Minor Impurities Identification
• High Resolution Image Processing

Innovation

• Sequential PatchCore
• Synthetic Data Training
• Fine-tuning with Real Data