Get Started
Home
Topics
Search
Library
Research questionHow can training-free visual anomaly detection catch local defects and global rule violations without category-specific modeling?Industrial images can be anomalous because of local texture damage or because objects violate expected counts, compositions, or arrangements. Detectors that handle one type often miss the other, especially when logical reasoning requires category-specific modeling.
AI
Computer Vision
Image & Video Processing
Machine Learning
Latest papersRecent research connected to this question, newest first.Training-Free Logical and Structural Anomaly Detection via Calibrated FusionThe source concerns a training-free detector that calibrates and fuses heterogeneous anomaly cues using normal images. Evidence is reported on the MVTec-LOCO and MVTec-AD benchmarks, including separate logical and structural anomaly results; deployment behavior beyond these benchmarks is not established.research paper · Sep 4, 2026
Related questions
How can frozen vision foundation models localize industrial defects from only a few examples?How can industrial visual anomaly detection generalize to small, weakly contrasted defects in imperfect factory images?How can industrial anomaly detection accelerate memory-bank construction and inference without sacrificing accuracy?How can industrial vision systems grade defect severity from imperfect detections while respecting ordinal levels and morphology?