Visible anomaly detection, an vital downside in pc imaginative and prescient, is often formulated as a one-class classification and segmentation job. The scholar-teacher (S-T) framework has proved to be efficient in fixing this problem. Nonetheless, earlier works based mostly on S-T solely empirically utilized constraints on regular information and fused multi-level info. On this research, we suggest an improved mannequin referred to as DeSTSeg, which integrates a pre-trained instructor community, a denoising pupil encoder-decoder, and a segmentation community into one framework. First, to strengthen the constraints on anomalous information, we introduce a denoising process that permits the coed community to study extra strong representations. From synthetically corrupted regular photos, we prepare the coed community to match the instructor community characteristic of the identical photos with out corruption. Second, to fuse the multi-level S-T options adaptively, we prepare a segmentation community with wealthy supervision from artificial anomaly masks, attaining a considerable efficiency enchancment. Experiments on the economic inspection benchmark dataset display that our technique achieves state-of-the-art efficiency, 98.6% on image-level ROC, 75.8% on pixel-level common precision, and 76.4% on instance-level common precision.
