Open set recognition through Monte Carlo dropout-based uncertainty. (23rd December 2021)

• Record Type: Journal Article
• Authors: Yin, Xiaojie; Hu, Qinghua; Schaefer, Gerald
• Is Part Of: International journal of bio-inspired computation. Volume 18:Number 4(2021)
• Page Start: 210
• Page End: 220
• Publication Date: 2021-12-23
• Subjects: open set recognition -- Monte Carlo dropout -- predictive uncertainty; Biologically-inspired computing -- Periodicals; Computational biology -- Periodicals; 572.0285
• Languages: English
• ISSNs: 1758-0366

Abstract

Open set recognition has received much attention in recent years. In this paper, we present a novel open set recognition method that is able to obtain improved recognition by applying Monte Carlo dropout to capture uncertainty in order to yield high quality predicted probabilities. Experimental results on six benchmark datasets show that our method gives better open set recognition performance than other state-of-the-art methods, with at least 6.4%, 3.9%, 2.9% and 1.0% performance increase in AUROC on the challenging object datasets CIFAR-10, CIFAR+10, CIFAR+50 and TinyImageNet respectively. We also perform an analysis on the benefits of combining predictive uncertainty with an EVT-based open set recognition model which indicates that Monte Carlo dropout-based uncertainty allows to obtain high quality predicted probabilities and to learn more accurate open set recognition scores. This, in turn, helps to reduce the overlap between known and unknown classes, thus making them more separable.