Research map / Video / Action recognition
TravelNet: Self-Supervised Physically Plausible Hand Motion Learning From Monocular Color Images
Zimeng Zhao, Xi Zhao, Yangang Wang
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 |
|---|---|---|---|
| add physical pose representation θ to the pose estimator (ablation) | TravelNet hand pose estimator · DO | AUC of PCK: .940 -> .962 | Improved |
| add physical pose representation θ to the pose estimator (ablation) | TravelNet hand pose estimator · STB | AUC of PCK: .890 -> .998 | Improved |
| add unified bone ratio γ extraction (ablation) | TravelNet hand pose estimator · RHD | AUC of PCK: .861 -> .903 | Improved |
| add self-supervised FKLayer term (ablation) | TravelNet hand pose estimator · STB | AUC of PCK: .956 -> .998 | Improved |
| add TravelNet over naive temporal smoothness of frame-wise estimates (ablation) | TravelNet · Noisy reconstruction σ=1.0 | average 3D joint error (mm): 32.57 -> 24.30 | Improved |
| add mean-shift clustering Fϵ in the encoder (ablation) | TravelNet · Noisy reconstruction σ=1.0 | average 3D joint error (mm): 28.05 -> 24.30 | Improved |
| add attention Fα in the encoder (ablation) | TravelNet · Noisy reconstruction σ=1.0 | average 3D joint error (mm): 26.68 -> 24.30 | Improved |
| add non-local attention in each learning block (ablation) | TravelNet · Noisy reconstruction σ=1.0 | average 3D joint error (mm): 25.36 -> 24.30 | Improved |
| add temporal smooth term LT in training (ablation) | TravelNet · Noisy reconstruction σ=1.0 | average 3D joint error (mm): 25.42 -> 24.30 | Improved |
| add collision penalty LC in training (ablation) | TravelNet · Noisy reconstruction σ=1.0 | average 3D joint error (mm): 24.31 -> 24.30 | No clear effect |
| add collision penalty LC in training (ablation) | TravelNet · DO sequences | average 3D joint error (mm): 17.05 -> 17.04 | No clear effect |
| use TravelNet fine-tuned encoder-decoder instead of frame-by-frame estimation | TravelNet · DO sequences (severe occlusion) | average 3D joint error (mm): 22.57 -> 17.04 | Improved |
Closest papers by meaning
- Consistent 3D Hand Reconstruction in Video via Self-Supervised Learning (TPAMI 2023 · similarity 0.88)
- PAD-Hand: Physics-Aware Diffusion for Hand Motion Recovery (CVPR 2026 · similarity 0.88)
- Model-Based 3D Hand Reconstruction via Self-Supervised Learning (CVPR 2021 · similarity 0.88)
- ComPose: When to Trust Hands for Object Pose Tracking (NeurIPS 2026 · similarity 0.87)
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video (CVPR 2024 · similarity 0.86)
- Hand-Centric Motion Refinement for 3D Hand-Object Interaction via Hierarchical Spatial-Temporal Modeling (AAAI 2024 · similarity 0.86)
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.
