Research map / 3D vision / Face and shape
Deep Virtual Markers for Articulated 3D Shapes
Hyomin Kim, Jungeon Kim, Jaewon Kam, Jaesik Park, Seungyong Lee
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use multiview aggregation (72 rendered depth views) instead of one-shot inference | Deep virtual markers · FAUST challenge intra-subject | error (cm): 1.417 -> 1.185 | Improved |
| use multiview aggregation instead of one-shot inference | Deep virtual markers · FAUST challenge inter-subject | error (cm): 2.495 -> 2.372 | Improved |
| use multiview aggregation instead of one-shot inference | Deep virtual markers · FAUST training pairs worst error, inter-subject | worst-pair error (cm): 4.56 -> 3.25 | Improved |
| use multiview aggregation instead of one-shot inference | Deep virtual markers · FAUST partial-to-partial inter-subject worst error | worst-pair error (cm): 16.06 -> 11.166 | Improved |
| use multiview aggregation instead of one-shot inference | Deep virtual markers · Ren et al. dataset S / F | normalized mean geodesic error (%): 3.1 -> 1.9 | Improved |
| use multiview aggregation instead of one-shot inference | Deep virtual markers · FAUST full-to-partial intra-subject | average error (cm): 2.41 -> 2.39 | No clear effect |
| train on soft geodesic labels instead of hard one-hot labels (ablation) | Deep virtual markers · Marker weight smoothness (qualitative) | visual smoothness | Improved |
| replace DHBC with deep virtual markers (multiview) | Full-to-full correspondence · FAUST training pairs intra-subject worst error | worst-pair error (cm): 9.98 -> 2.76 | Improved |
| replace DHBC with deep virtual markers (multiview) | Partial-to-partial correspondence · FAUST P2P intra-subject | average error (cm): 9.51 -> 3.40 | Improved |
| replace DHBC with deep virtual markers (oneshot) | Full-to-full correspondence · FAUST training pairs intra-subject average error | average error (cm): 2.00 -> 2.26 | Hurt |
| train on SCAPE only and test on FAUST (limited shape variety), versus F / F (ablation) | 3D-CODED · Ren et al. dataset | normalized mean geodesic error (%): 2.5 -> 33.0 | Hurt |
Closest papers by meaning
- 3D Human Mesh Estimation From Virtual Markers (CVPR 2023 · similarity 0.89)
- VMarker-Pro: Probabilistic 3D Human Mesh Estimation From Virtual Markers (TPAMI 2025 · similarity 0.87)
- DeepTag: A General Framework for Fiducial Marker Design and Detection (TPAMI 2022 · similarity 0.85)
- Towards 3D Human Pose Estimation in the Wild: A Weakly-Supervised Approach (ICCV 2017 · similarity 0.85)
- ViSER: Video-Specific Surface Embeddings for Articulated 3D Shape Reconstruction (NeurIPS 2021 · similarity 0.85)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
