Gesture-Aware Pretraining and Token Fusion for 3D Hand Pose Estimation

• URL: http://arxiv.org/abs/2603.17396v1
• Date: Wed, 18 Mar 2026 06:18:46 GMT
• Title: Gesture-Aware Pretraining and Token Fusion for 3D Hand Pose Estimation
• Abstract summary: We show that gesture semantics can serve as a powerful inductive bias for 3D pose estimation.<n>We present a two-stage framework: gesture-aware pretraining that learns an informative embedding space.<n>Experiments on InterHand2.6M demonstrate that gesture-aware pretraining consistently improves single-hand accuracy.
• Score: 3.0979829079231447
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Estimating 3D hand pose from monocular RGB images is fundamental for applications in AR/VR, human-computer interaction, and sign language understanding. In this work we focus on a scenario where a discrete set of gesture labels is available and show that gesture semantics can serve as a powerful inductive bias for 3D pose estimation. We present a two-stage framework: gesture-aware pretraining that learns an informative embedding space using coarse and fine gesture labels from InterHand2.6M, followed by a per-joint token Transformer guided by gesture embeddings as intermediate representations for final regression of MANO hand parameters. Training is driven by a layered objective over parameters, joints, and structural constraints. Experiments on InterHand2.6M demonstrate that gesture-aware pretraining consistently improves single-hand accuracy over the state-of-the-art EANet baseline, and that the benefit transfers across architectures without any modification.

Related papers

• TransHands: Repurposing Human Pose Encoders as Hand Pose Encoders [3.75886080255807]
  We introduce TransHands, a backbone-agnostic transfer learning framework that enables pre-trained human motion encoders to be effectively adapted for 3D hand pose estimation.<n>We evaluate TransHands across four state-of-the-art motion modeling architectures, including transformer-based, graph-based, and frequency- domain models.
  arXiv  Detail & Related papers  (2026-08-23T10:17:09Z)
• AffordGrasp: Cross-Modal Diffusion for Affordance-Aware Grasp Synthesis [50.793806818677716]
  AffordGrasp produces physically stable and semantically faithful human grasps with high precision.<n>AffordGrasp integrates an affordance-aware latent representation of hand poses with a dual-conditioning diffusion process.<n>We evaluate AffordGrasp across four instruction-augmented benchmarks derived from HO-3D, OakInk, GRAB, and AffordPose.
  arXiv  Detail & Related papers  (2026-03-09T06:56:35Z)
• Learning to Align and Refine: A Foundation-to-Diffusion Framework for Occlusion-Robust Two-Hand Reconstruction [50.952228546326516]
  Two-hand reconstruction from monocular images faces persistent challenges due to complex and dynamic hand postures.<n>Existing approaches struggle with such alignment issues, often resulting in misalignment and penetration artifacts.<n>We propose a dual-stage Foundation-to-Diffusion framework that precisely align 2D prior guidance from vision foundation models.
  arXiv  Detail & Related papers  (2025-03-22T14:42:27Z)
• Learning Interaction-aware 3D Gaussian Splatting for One-shot Hand Avatars [47.61442517627826]
  We propose to create animatable avatars for interacting hands with 3D Gaussian Splatting (GS) and single-image inputs. Our proposed method is validated via extensive experiments on the large-scale InterHand2.6M dataset.
  arXiv  Detail & Related papers  (2024-10-11T14:14:51Z)
• DICE: End-to-end Deformation Capture of Hand-Face Interactions from a Single Image [98.29284902879652]
  We present DICE, the first end-to-end method for Deformation-aware hand-face Interaction reCovEry from a single image.<n>It features disentangling the regression of local deformation fields and global mesh locations into two network branches.<n>It achieves state-of-the-art performance on a standard benchmark and in-the-wild data in terms of accuracy and physical plausibility.
  arXiv  Detail & Related papers  (2024-06-26T00:08:29Z)
• HOISDF: Constraining 3D Hand-Object Pose Estimation with Global Signed Distance Fields [96.04424738803667]
  HOISDF is a guided hand-object pose estimation network. It exploits hand and object SDFs to provide a global, implicit representation over the complete reconstruction volume. We show that HOISDF achieves state-of-the-art results on hand-object pose estimation benchmarks.
  arXiv  Detail & Related papers  (2024-02-26T22:48:37Z)
• CLIP-Hand3D: Exploiting 3D Hand Pose Estimation via Context-Aware Prompting [38.678165053219644]
  We make one of the first attempts to propose a novel 3D hand pose estimator from monocular images, dubbed as CLIP-Hand3D. We maximize semantic consistency for a pair of pose-text features following a CLIP-based contrastive learning paradigm. Experiments on several public hand benchmarks show that the proposed model attains a significantly faster inference speed.
  arXiv  Detail & Related papers  (2023-09-28T03:40:37Z)
• Monocular 3D Reconstruction of Interacting Hands via Collision-Aware Factorized Refinements [96.40125818594952]
  We make the first attempt to reconstruct 3D interacting hands from monocular single RGB images. Our method can generate 3D hand meshes with both precise 3D poses and minimal collisions.
  arXiv  Detail & Related papers  (2021-11-01T08:24:10Z)
• 3D Hand Pose and Shape Estimation from RGB Images for Improved Keypoint-Based Hand-Gesture Recognition [25.379923604213626]
  This paper presents a keypoint-based end-to-end framework for the 3D hand and pose estimation. It is successfully applied to the hand-gesture recognition task as a study case.
  arXiv  Detail & Related papers  (2021-09-28T17:07:43Z)
• Learning to Disambiguate Strongly Interacting Hands via Probabilistic Per-pixel Part Segmentation [84.28064034301445]
  Self-similarity, and the resulting ambiguities in assigning pixel observations to the respective hands, is a major cause of the final 3D pose error. We propose DIGIT, a novel method for estimating the 3D poses of two interacting hands from a single monocular image. We experimentally show that the proposed approach achieves new state-of-the-art performance on the InterHand2.6M dataset.
  arXiv  Detail & Related papers  (2021-07-01T13:28:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.