CRAS is a unified multi-class industrial anomaly detection method whose center-aware residual features and distance-guided synthetic anomalies reach 98.3% image-level AUROC on MVTec AD.
An anomaly feature-editing- based adversarial network for texture defect visual inspection,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Center-aware Residual Anomaly Synthesis for Multi-class Industrial Anomaly Detection
CRAS is a unified multi-class industrial anomaly detection method whose center-aware residual features and distance-guided synthetic anomalies reach 98.3% image-level AUROC on MVTec AD.