Please use this identifier to cite or link to this item:
Scopus Web of Science® Altmetric
Type: Journal article
Title: Uncertainty for identifying open-set errors in visual object detection
Author: Miller, D.
Sunderhauf, N.
Milford, M.
Dayoub, F.
Citation: IEEE Robotics and Automation Letters, 2022; 7(1):215-222
Publisher: Institute of Electrical and Electronics Engineers (IEEE)
Issue Date: 2022
ISSN: 2377-3766
Statement of
Dimity Miller, Niko Sünderhauf, Michael Milford, Feras Dayoub
Abstract: Deployed into an open world, object detectors are prone to open-set errors, false positive detections of object classes not present in the training dataset.We propose GMM-Det, a real-time method for extracting epistemic uncertainty from object detectors to identify and reject open-set errors. GMM-Det trains the detector to produce a structured logit space that is modelled with class-specific Gaussian Mixture Models. At test time, open-set errors are identified by their low log-probability under all Gaussian Mixture Models. We test two common detector architectures, Faster R-CNN and RetinaNet, across three varied datasets spanning robotics and computer vision. Our results show that GMM-Det consistently outperforms existing uncertainty techniques for identifying and rejecting open-set detections, especially at the low-error-rate operating point required for safety-critical applications. GMM-Det maintains object detection performance, and introduces only minimal computational overhead. We also introduce a methodology for converting existing object detection datasets into specific open-set datasets to evaluate open-set performance in object detection.
Keywords: detection; segmentation and categorization; deep learning for visual perception
Rights: © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
DOI: 10.1109/lra.2021.3123374
Grant ID:
Appears in Collections:Physics publications

Files in This Item:
There are no files associated with this item.

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.