Vis enkel innførsel

dc.contributor.authorLi, Haoyuan
dc.contributor.authorDong, Haoye
dc.contributor.authorJia, Hanchao
dc.contributor.authorHuang, Dong
dc.contributor.authorKampffmeyer, Michael Christian
dc.contributor.authorLin, Liang
dc.contributor.authorLiang, Xiaodan
dc.date.accessioned2024-02-16T13:13:48Z
dc.date.available2024-02-16T13:13:48Z
dc.date.issued2024-01-15
dc.description.abstractMulti-person 3D mesh recovery from videos is a critical first step towards automatic perception of group behavior in virtual reality, physical therapy and beyond. However, existing approaches rely on multi-stage paradigms, where the person detection and tracking stages are performed in a multi-person setting, while temporal dynamics are only modeled for one person at a time. Consequently, their performance is severely limited by the lack of inter-person interactions in the spatial-temporal mesh recovery, as well as by detection and tracking defects. To address these challenges, we propose the Coordinate transFormer (Coord-Former) that directly models multi-person spatial-temporal relations and simultaneously performs multi-mesh recovery in an end-to-end manner Instead of partitioning the feature map into coarse-scale patch-wise tokens, CoordFormer leverages a novel Coordinate-Aware Attention to preserve pixel-level spatial-temporal coordinate information. Additionally, we propose a simple, yet effective Body Center Attention mechanism to fuse position information. Extensive experiments on the 3DPW dataset demonstrate that CoordFormer significantly improves the state-of-the-art, outperforming the previously best results by 4.2%, 8.8% and 4.7% according to the MPJPE, PAMPJPE, and PVE metrics, respectively, while being 40% faster than recent video-based approaches. The released code can be found at https://github.com/Li-Hao-yuan/CoordFormer.en_US
dc.identifier.citationLi H, Dong H, Jia, Huang D, Kampffmeyer MC, Lin L, Liang X. Coordinate Transformer: Achieving Single-stage Multi-person Mesh Recovery from Videos. IEEE International Conference on Computer Vision (ICCV). 2023en_US
dc.identifier.cristinIDFRIDAID 2185859
dc.identifier.doi10.1109/ICCV51070.2023.00803
dc.identifier.issn1550-5499
dc.identifier.issn2380-7504
dc.identifier.urihttps://hdl.handle.net/10037/32953
dc.language.isoengen_US
dc.publisherIEEEen_US
dc.relation.journalIEEE International Conference on Computer Vision (ICCV)
dc.relation.projectIDNorges forskningsråd: 309439en_US
dc.relation.projectIDNorges forskningsråd: 315029en_US
dc.rights.accessRightsopenAccessen_US
dc.rights.holderCopyright 2023 The Author(s)en_US
dc.titleCoordinate Transformer: Achieving Single-stage Multi-person Mesh Recovery from Videosen_US
dc.type.versionacceptedVersionen_US
dc.typeJournal articleen_US
dc.typeTidsskriftartikkelen_US
dc.typePeer revieweden_US


Tilhørende fil(er)

Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel