AbstractAnomaly detection can be conceived either through generative modelling of regular training data or by discriminating with respect to negative data. These two approaches exhibit different failure modes. Consequently, hybrid algorithms present an attractive research goal. Unfortunately, dense anomaly requires translational equivariance and very large input resolutions. requirements disqua...

One of the challenges in pattern recognition is open set recognition. Compared with closed recognition, needs to reduce not only empirical risk, but also space and reduction these two risks corresponds classifying known classes identifying unknown respectively. How risk key This paper explores origin by analyzing distribution features. On this basis, spatial location constraint prototype loss f...