FAIR: Frequency-aware Image Restoration for Industrial Visual Anomaly Detection
This work addresses anomaly detection in industrial visual inspection, offering an incremental improvement by leveraging frequency biases to enhance detection accuracy and efficiency.
The paper tackles the trade-off between normal reconstruction fidelity and abnormal reconstruction distinguishability in image reconstruction-based anomaly detection for industrial visual inspection by proposing FAIR, a self-supervised image restoration task that restores images from high-frequency components, achieving state-of-the-art performance with higher efficiency on various defect detection datasets.
Image reconstruction-based anomaly detection models are widely explored in industrial visual inspection. However, existing models usually suffer from the trade-off between normal reconstruction fidelity and abnormal reconstruction distinguishability, which damages the performance. In this paper, we find that the above trade-off can be better mitigated by leveraging the distinct frequency biases between normal and abnormal reconstruction errors. To this end, we propose Frequency-aware Image Restoration (FAIR), a novel self-supervised image restoration task that restores images from their high-frequency components. It enables precise reconstruction of normal patterns while mitigating unfavorable generalization to anomalies. Using only a simple vanilla UNet, FAIR achieves state-of-the-art performance with higher efficiency on various defect detection datasets. Code: https://github.com/liutongkun/FAIR.