Topology-preserving transfer learning for weakly-supervised anomaly detection and segmentation

Shenxing Wei, Xing Wei, Muhammad Rifki Kurniawan, Zhiheng Ma, Yihong Gong

Abstract: Highlights•Naïve transfer techniques lead to feature space collapse in anomaly detection and segmentation area.•Topology-preserve constraint can mitigate collapse conspicuously.•Significant improvement over previous methods on BTAD and CIFAR100/10 datasets.