Research map / 3D vision / Face and shape
TokenHand: Discrete Token Representation for Efficient Hand Mesh Reconstruction
Xinguo He, Yixin Shen, Rahul Chaudhari
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 |
|---|---|---|---|
| increase regressor layers from 1 to 4 with progressive x2 token counts (ablation) | TokenHand decoder · FreiHAND test | PA-MPJPE (mm): 6.4 -> 5.7 | Improved |
| use non-progressive token counts [48,128,256,512] instead of [48,97,194,389] (ablation) | TokenHand decoder · FreiHAND test | PA-MPJPE (mm): 5.7 -> 5.9 | Hurt |
| reduce feature dimensions toward 16 instead of 32 (ablation) | TokenHand decoder · FreiHAND test | PA-MPJPE (mm): 5.7 -> 5.9 | Hurt |
| increase MetaFormer blocks per regressor layer from 1 to 3 (ablation) | TokenHand decoder · FreiHAND test | PA-MPJPE (mm): 5.7 -> 5.8 | No clear effect |
| change codebook size (384x512, 640x512 vs 512x512) (ablation) | TokenHand Stage I · FreiHAND | PA-MPJPE (mm): 4.9 -> 5.1 (640x512) | No clear effect |
| switch backbone from ResNet50 to FastViT-MA36 (ablation) | TokenHand · FreiHAND test | PA-MPJPE (mm): 6.7 -> 5.7 | Improved |
| use global pooled feature as token feature instead of keypoint-guided sampling on upsampled features (ablation) | TokenHand token generator · FreiHAND test | PA-MPJPE (mm): 5.7 -> 6.2 | Hurt |
| add 4x upsampling before keypoint-guided sampling (ablation) | TokenHand token generator · FreiHAND test | PA-MPJPE (mm): 6.0 -> 5.7 | Improved |
| add convolutional enhancement of upsampled features (ablation) | TokenHand token generator · FreiHAND test | PA-MPJPE (mm): 5.7 -> 5.7 | No clear effect |
| replace keypoint-based sampling with coarse-mesh sampling (ablation) | TokenHand token generator · FreiHAND test | PA-MPJPE (mm): 5.7 -> 5.8 | No clear effect |
| discrete token classification with lightweight decoder instead of transformer regression | hand mesh reconstruction · FreiHAND test | #Params (decoder): 98M -> 3.0M vs MeshGraphormer | Improved |
| use TokenHand instead of FastViT | FastViT-MA36 backbone · FreiHAND test | FPS: 84 -> 65 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- MobRecon — MobRecon: Mobile-Friendly Hand Mesh Reconstruction From Monocular Image (CVPR 2022) · better on 1 of 2 shared results
- H2ONet — H2ONet: Hand-Occlusion-and-Orientation-Aware Network for Real-Time 3D Hand Mesh Reconstruction (CVPR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Simple Baseline for Efficient Hand Mesh Reconstruction (CVPR 2024 · similarity 0.91)
- HandBooster: Boosting 3D Hand-Mesh Reconstruction by Conditional Synthesis and Sampling of Hand-Object Interactions (CVPR 2024 · similarity 0.90)
- Fine-Grained Multi-View Hand Reconstruction Using Inverse Rendering (AAAI 2024 · similarity 0.90)
- HandOS: 3D Hand Reconstruction in One Stage (CVPR 2025 · similarity 0.89)
- Reconstructing Hands in 3D with Transformers (CVPR 2024 · similarity 0.88)
- SRHand: Super-Resolving Hand Images and 3D Shapes via View/Pose-aware Neural Image Representations and Explicit Meshes (NeurIPS 2025 · similarity 0.88)
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.
