Per-frame mAP Prediction for Continuous Performance Monitoring of Object Detection during Deployment

Simple item page
dc.contributor.authorRahman, Q.M.
dc.contributor.authorSunderhauf, N.
dc.contributor.authorDayoub, F.
dc.contributor.conferenceIEEE Winter Conference on Applications of Computer Vision Workshops (WACVW) (5 Jan 2021 - 9 Jan 2021 : Waikola, HI, USA)
dc.date.issued2021
dc.description.abstractPerformance monitoring of object detection is crucial for safety-critical applications such as autonomous vehicles that operate under varying and complex environmental conditions. Currently, object detectors are evaluated using summary metrics based on a single dataset that is assumed to be representative of all future deployment conditions. In practice, this assumption does not hold, and the performance fluctuates as a function of the deployment conditions. To address this issue, we propose an introspection approach to performance monitoring during deployment without the need for ground truth data. We do so by predicting when the per-frame mean average precision drops below a critical threshold using the detector’s internal features. We quantitatively evaluate and demonstrate our method’s ability to reduce risk by trading off making an incorrect decision by raising the alarm and absenting from detection.
dc.description.statementofresponsibilityQuazi Marufur Rahman, Niko Sunderhauf, Feras Dayoub
dc.identifier.citationProceedings of the IEEE Winter Conference on Applications of Computer Vision Workshops (WACVW 2021), 2021, pp.152-160
dc.identifier.doi10.1109/WACVW52041.2021.00021
dc.identifier.isbn9781665419673
dc.identifier.issn2572-4398
dc.identifier.orcidDayoub, F. [0000-0002-4234-7374]
dc.identifier.urihttps://hdl.handle.net/2440/134823
dc.language.isoen
dc.publisherIEEE
dc.publisher.placeonline
dc.relation.granthttp://purl.org/au-research/grants/arc/CE140100016
dc.relation.ispartofseriesIEEE Winter Conference on Applications of Computer Vision Workshops
dc.rights©2021 IEEE
dc.source.urihttps://ieeexplore.ieee.org/xpl/conhome/9407778/proceeding
dc.subjectMeasurement; Deep learning; Computer vision; Conferences; Object detection; Detectors; Feature extraction
dc.titlePer-frame mAP Prediction for Continuous Performance Monitoring of Object Detection during Deployment
dc.typeConference paper
pubs.publication-statusPublished

Files

Collections

Computer Science publications