Research map / Video / Action recognition
DisentangledTMR: Privacy-Preserving Skeleton Motion Retargeting via Factorized Transformers
Thomas Carr, Depeng Xu, Shuhan Yuan, Aidong Lu
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 dual encoders with factorized decoder instead of single-encoder PMR | DisentangledTMR · NTU60 X-View | pre-trained AR: 35.7 -> 75.8 | Improved |
| use DisentangledTMR instead of DMR | DMR · NTU60 X-View | pre-trained RI: 25.7 -> 18.1 | Improved |
| use DisentangledTMR instead of PMR | PMR · NTU60 X-View | reconstruction MSE: 0.050 -> 0.030 | Improved |
| use DisentangledTMR instead of DMR for bone-length consistency | DMR · NTU60 X-View | bone length error: 0.029 -> 0.016 | Improved |
| add Stage 2 and Stage 3 training after Stage 1 (ablation) | DisentangledTMR · NTU60 X-View | pre-trained AR: 57.3 -> ~82 | Improved |
| add Stages 2 and 3 after Stage 1 (privacy) (ablation) | DisentangledTMR · NTU60 X-View | pre-trained RI: 15.3 | Hurt |
| raise partial-retargeting ratio beta from 0.20 to 0.30 | DisentangledTMR · NTU60 X-View | pre-trained AR: 76.8 -> 42.8 | Hurt |
| raise beta from 0.00 to 0.20 | DisentangledTMR · NTU60 X-View | pre-trained RI: 75.4 -> 17.3 | Improved |
| raise beta to 1.00 (full retargeting) | DisentangledTMR · NTU60 X-View | pre-trained AR: 2.0 | Hurt |
| raise beta from 0.00 to 0.20 on NTU120 | DisentangledTMR · NTU120 | pre-trained RI: 67.6 -> 39.4 | Improved |
| raise beta from 0.00 to 0.20 on ETRI | DisentangledTMR · ETRI | pre-trained RI: 45.6 -> 35.7 | Improved |
| use the output-level supervision suite (NTU60) vs earlier checkpoints | DisentangledTMR · NTU60 vs NTU120/ETRI | pre-trained AR: 76.8%->42.8% | Mixed |
Closest papers by meaning
- Motif Guided Graph Transformers with Combinatorial Skeleton Prototype Learning for Skeleton-Based Person Re-Identification (AAAI 2025 · similarity 0.86)
- MaskCLR: Attention-Guided Contrastive Learning for Robust Action Representation Learning (CVPR 2024 · similarity 0.86)
- Less is More: Compact-Token Masked Feature Learning for Skeleton Representation Learning (NeurIPS 2026 · similarity 0.85)
- SAMoR: Motion Modelling for Articulated Objects of Any Skeleton and Topology (NeurIPS 2026 · similarity 0.84)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.84)
- Skeleton-based Action Recognition with Non-linear Dependency Modeling and Hilbert-Schmidt Independence Criterion (AAAI 2025 · similarity 0.84)
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.
