We are very glad to announce that our paper entitled “Nearest neighbors distance ratio open-set classifier” has been accepted in the very prestigious Springer Machine Learning journal.
In this work, the authors propose a novel multiclass classifier for open-set recognition setups, in which there are no a priori training samples for some classes. The proposed Open-Set NN (OSNN) method, which extends upon the Nearest-Neighbor (NN) classifier, incorporates the ability of recognizing samples belonging to classes that are unknown at training time. The method is validated using large benchmarks with different open-set recognition regimes showing that OSNN significantly outperforms its counterparts in the literature.
A pre-print version of the paper is available here.