Research map / 3D vision / Face and shape
Im2Hands: Learning Attentive Implicit Representation of Interacting Two-Hand Shapes
Jihyun Lee, Minhyuk Sung, Honggyu Choi, Tae-Kyun Kim
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 Im2Hands instead of HALO (modified) with RGB image and keypoints | Two-hand reconstruction · InterHand2.6M | IoU: 75.8 -> 77.8 | Improved |
| Add keypoint refinement K to DIGIT-predicted keypoints | DIGIT · InterHand2.6M | MPJPE: 16.75 -> 10.70 | Improved |
| Add keypoint refinement K to IntagHand-predicted keypoints | IntagHand keypoint predictions · InterHand2.6M | MPJPE: 10.13 -> 9.68 | Improved |
| Use Im2Hands instead of HALO with DIGIT keypoints | DIGIT + HALO · InterHand2.6M single-image | IoU: 45.1 -> 59.4 | Improved |
| Add keypoint refinement K (ablation) | Im2Hands initial occupancy I with DIGIT keypoints · InterHand2.6M single-image | IoU: 45.6 -> 55.4 | Improved |
| Add occupancy refinement R (ablation) | K + I with DIGIT keypoints · InterHand2.6M single-image | IoU: 55.4 -> 59.4 | Improved |
| Add RGB image conditioning (ablation) | Initial occupancy I · InterHand2.6M ground-truth keypoints | IoU: 74.7 -> 77.2 | Improved |
| Use query-image attention instead of pixel-aligned image features (ablation) | Initial occupancy I · InterHand2.6M ground-truth keypoints | IoU: 75.8 -> 77.2 | Improved |
| Condition refined occupancy on initial occupancy probabilities (ablation) | I + R · InterHand2.6M ground-truth keypoints | IoU: 67.0 -> 77.8 | Improved |
| Use feature cloud conversion in occupancy refinement (ablation) | I + R · InterHand2.6M ground-truth keypoints | IoU: 77.4 -> 77.8 | Improved |
| Use context latent extraction instead of separate per-hand global latents (ablation) | I + R · InterHand2.6M ground-truth keypoints | IoU: 77.6 -> 77.8 | Improved |
| Use 256 rather than 512 sampled surface points (ablation) | Im2Hands occupancy refinement · InterHand2.6M ground-truth keypoints | IoU: 77.4 vs 77.8 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports Im2Hands as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- FourierHandFlow: Neural 4D Hand Representation Using Fourier Query Flow (NeurIPS 2023 · better on 2 of 2 shared results)
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
- SRHand: Super-Resolving Hand Images and 3D Shapes via View/Pose-aware Neural Image Representations and Explicit Meshes (NeurIPS 2025 · similarity 0.90)
- InterHandGen: Two-Hand Interaction Generation via Cascaded Reverse Diffusion (CVPR 2024 · similarity 0.90)
- Learning Explicit Contact for Implicit Reconstruction of Hand-Held Objects from Monocular Images (AAAI 2024 · similarity 0.89)
- BOTH2Hands: Inferring 3D Hands from Both Text Prompts and Body Dynamics (CVPR 2024 · similarity 0.89)
- FoundHand: Large-Scale Domain-Specific Learning for Controllable Hand Image Generation (CVPR 2025 · similarity 0.89)
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video (CVPR 2024 · 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.
