SuperADD enhances SuperAD with DINOv3 backbone, overlapping patches, augmentations, better memory subsampling, and morphological closing to achieve higher F1 scores on MVTec AD 2 without class-specific tuning.
Training-free indus- trial defect generation with diffusion models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SuperADD: Training-free Class-agnostic Anomaly Segmentation -- CVPR 2026 VAND 4.0 Workshop Challenge Industrial Track
SuperADD enhances SuperAD with DINOv3 backbone, overlapping patches, augmentations, better memory subsampling, and morphological closing to achieve higher F1 scores on MVTec AD 2 without class-specific tuning.