Unsupervised defect detection with patch-aware mutual reasoning network in image data

Abstract

Current defect detection studies in the industrial fields mainly adopt supervised strategies, which require large amounts of annotated defective samples to achieve superior results. However, it is hard to meet such large-scale data requirements in the actual industrial scenarios where defects commonly exhibit intensive scarcity and huge intraclass variation. To address the above limitations, we propose a novel Patch-aware Mutual Reasoning Network (PMRN) that utilizes only the prior knowledge of non-defective samples for defect detection. Concretely, a patch-aware mutual reasoning module and a spatial shuffle perception module are devised to reason mutual dependencies and explore dislocations relationships. Besides, an adaptive soft gated anomaly measurement function is developed to calculate reconstruction deviations, which can soft control the information flow according to the complexity of the current scenario. Extensive experimental results (five benchmark datasets: average AUC of 0.951) demonstrate that the PMRN can accurately detect defects with complex backgrounds and weak textures using only non-defective data for training, thus relieving the dependence on large-scale annotated defective samples. Future work will focus on the unsupervised detection of tiny defects.