AnoRefiner: Anomaly-Aware Group-Wise Refinement for Zero-Shot Industrial Anomaly Detection

TL;DR

AnoRefiner is a novel anomaly-aware refinement method that enhances zero-shot industrial anomaly detection by improving patch-level maps to pixel-level accuracy using anomaly score maps and progressive training, boosting detection performance.

Contribution

It introduces a plug-in anomaly-aware refiner with an anomaly refinement decoder and a progressive group-wise training strategy, addressing the gap between synthetic and real anomalies in ZSAD.

Findings

Boosts pixel-AP metrics by up to 5.2% across models.

Effectively improves fine-grained anomaly detection in industrial images.

Compatible with various ZSAD methods and applicable to real-world datasets.

Abstract

Zero-shot industrial anomaly detection (ZSAD) methods typically yield coarse anomaly maps as vision transformers (ViTs) extract patch-level features only. To solve this, recent solutions attempt to predict finer anomalies using features from ZSAD, but they still struggle to recover fine-grained anomalies without missed detections, mainly due to the gap between randomly synthesized training anomalies and real ones. We observe that anomaly score maps exactly provide complementary spatial cues that are largely absent from ZSAD's image features, a fact overlooked before. Inspired by this, we propose an anomaly-aware refiner (AnoRefiner) that can be plugged into most ZSAD models and improve patch-level anomaly maps to the pixel level. First, we design an anomaly refinement decoder (ARD) that progressively enhances image features using anomaly score maps, reducing the reliance on synthetic anomaly data. Second, motivated by the mass production paradigm, we propose a progressive group-wise test-time training (PGT) strategy that trains ARD in each product group for the refinement process in the next group, while staying compatible with any ZSAD method. Experiments on the MVTec AD and VisA datasets show that AnoRefiner boosts various ZSAD models by up to a 5.2\% gain in pixel-AP metrics, which can also be directly observed in many visualizations. The code will be available at https://github.com/HUST-SLOW/AnoRefiner.