Open-set Recognition via Augmentation-based Similarity Learning

Sepideh Esmaeilpour, Lei Shu, Bing Liu

The primary assumption of conventional supervised learning or classification is that the test samples are drawn from the same distribution as the training samples, which is called…