ijbic21.pdf (953.44 kB)
Open set recognition through Monte Carlo dropout-based uncertainty
journal contribution
posted on 2022-01-07, 14:22 authored by Xiaojie Yin, Qinghua Hu, Gerald SchaeferGerald SchaeferOpen 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 stateof-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.
Funding
National Natural Science Foundation of China under Grants 61925602 and 61732011
History
School
- Science
Department
- Computer Science
Published in
International Journal of Bio-Inspired ComputationVolume
18Issue
4Pages
210-220Publisher
InderscienceVersion
- AM (Accepted Manuscript)
Rights holder
© IndersciencePublisher statement
This paper was accepted for publication in the journal International Journal of Bio-Inspired Computation and the definitive published version is available at https://doi.org/10.1504/IJBIC.2021.119982Acceptance date
2021-02-22Publication date
2021-12-23Copyright date
2021ISSN
1758-0366eISSN
1758-0374Publisher version
Language
- en