Full metadata record

DC Field Value Language
dc.contributor.authorDogyun Park-
dc.contributor.authorKim, Suhyun-
dc.date.accessioned2024-01-12T02:44:53Z-
dc.date.available2024-01-12T02:44:53Z-
dc.date.created2023-11-26-
dc.date.issued2023-10-
dc.identifier.urihttps://pubs.kist.re.kr/handle/201004/76371-
dc.identifier.urihttps://iccv2023.thecvf.com/main.conference.program-107.php-
dc.description.abstractAssessing the fidelity and diversity of the generative model is a difficult but important issue for technological advancement. So, recent papers have introduced k-Nearest Neighbor (kNN) based precision-recall metrics to break down the statistical distance into fidelity and diversity. While they provide an intuitive method, we thoroughly analyze these metrics and identify oversimplified assumptions and undesirable properties of kNN that result in unreliable evaluation, such as susceptibility to outliers and insensitivity to distributional changes. Thus, we propose novel metrics, P-precision and Precall (PP&PR), based on a probabilistic approach that address the problems. Through extensive investigations on toy experiments and state-of-the-art generative models, we show that our PP&PR provide more reliable estimates for comparing fidelity and diversity than the existing metrics. The codes are available at https://github.com/kdst-team/ Probablistic_precision_recall.-
dc.publisherThe Computer Vision Foundation-
dc.titleProbabilistic Precision and Recall Towards Reliable Evaluation of Generative Models-
dc.typeConference-
dc.identifier.doi10.1109/ICCV51070.2023.01839-
dc.description.journalClass1-
dc.identifier.bibliographicCitation2023 International Conference on Computer Vision (ICCV 2023)-
dc.citation.title2023 International Conference on Computer Vision (ICCV 2023)-
dc.citation.conferencePlaceFR-
dc.citation.conferencePlaceParis Convention Centre-
dc.citation.conferenceDate2023-10-02-
dc.relation.isPartOfInternational Conference on Computer Vision (ICCV 2023)-
dc.identifier.wosid001169500504060-
dc.identifier.scopusid2-s2.0-85185872875-
Appears in Collections:
KIST Conference Paper > 2023
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML

qrcode

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

BROWSE