DC Field | Value | Language |
---|---|---|
dc.contributor.author | Dong Hun Ryou | - |
dc.contributor.author | KIM, YOUWANG | - |
dc.contributor.author | Oh, Tae-Hyun | - |
dc.date.accessioned | 2023-03-09T02:00:27Z | - |
dc.date.available | 2023-03-09T02:00:27Z | - |
dc.date.created | 2023-03-02 | - |
dc.date.issued | 2024-02 | - |
dc.identifier.issn | 0178-2789 | - |
dc.identifier.uri | https://oasis.postech.ac.kr/handle/2014.oak/116949 | - |
dc.description.abstract | We present a rank statistic adaptive multi-stage pruning method to find lightweight neural networks for 3D human mesh recovery while minimizing accuracy drop. We observe that some feature maps often have prominent low-rank patterns regardless of input human images. Furthermore, even after pruning, feature channels that should have been pruned according to pruning criteria frequently re-appear in test time. From these observations, we design rank statistic adaptive multi-stage pruning; thereby, we can prune more filters with recovering mesh reconstruction accuracy. We demonstrate that, for DenseNet-121, 60.0% of parameters and 67.9% of FLOPs are saved while maintaining comparable accuracy to that of the original full model. This is a notable improvement compared to the competing method based on the L1 filter pruning, where the error is increased by 17.55% at the same pruning rate. | - |
dc.language | English | - |
dc.publisher | Springer Verlag | - |
dc.relation.isPartOf | Visual Computer | - |
dc.title | Multi-stage Adaptive Rank Statistic Pruning for Lightweight Human 3D Mesh Recovery Model | - |
dc.type | Article | - |
dc.identifier.doi | 10.1007/s00371-023-02798-x | - |
dc.type.rims | ART | - |
dc.identifier.bibliographicCitation | Visual Computer, v.40, no.2, pp.535 - 543 | - |
dc.identifier.wosid | 000945767400001 | - |
dc.citation.endPage | 543 | - |
dc.citation.number | 2 | - |
dc.citation.startPage | 535 | - |
dc.citation.title | Visual Computer | - |
dc.citation.volume | 40 | - |
dc.contributor.affiliatedAuthor | KIM, YOUWANG | - |
dc.contributor.affiliatedAuthor | Oh, Tae-Hyun | - |
dc.identifier.scopusid | 2-s2.0-85149458738 | - |
dc.description.journalClass | 1 | - |
dc.description.journalClass | 1 | - |
dc.description.isOpenAccess | N | - |
dc.type.docType | Article | - |
dc.subject.keywordAuthor | Human mesh recovery | - |
dc.subject.keywordAuthor | Lightweight neural networks | - |
dc.subject.keywordAuthor | Low-rank matrix | - |
dc.subject.keywordAuthor | Pruning | - |
dc.relation.journalWebOfScienceCategory | Computer Science, Software Engineering | - |
dc.description.journalRegisteredClass | scie | - |
dc.description.journalRegisteredClass | scopus | - |
dc.relation.journalResearchArea | Computer Science | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
library@postech.ac.kr Tel: 054-279-2548
Copyrights © by 2017 Pohang University of Science ad Technology All right reserved.