Few Shot Part Segmentation Reveals Compositional Logic for Industrial Anomaly Detection

arXiv:2312.13783v2 Announce Type: replace-cross
Abstract: Logical anomalies (LA) refer to data violating underlying logical constraints e.g., the quantity, arrangement, or composition of components within an image. Detecting accurately such anomalies requires models to reason about various component types through segmentation. However, curation of pixel-level annotations for semantic segmentation is both time-consuming and expensive. Although there are some prior few-shot or unsupervised co-part segmentation algorithms, they often fail on images with industrial object. These images have components with similar textures and …

abstract annotations anomaly anomaly detection arxiv components constraints cs.ai cs.cv cs.lg curation data detection however image industrial logic part pixel reason segmentation semantic through type types