DexFlow: A Unified Approach for Dexterous Hand Pose Retargeting and Interaction

TL;DR

DexFlow introduces a novel data transformation pipeline that enhances the accuracy, naturalness, and diversity of dexterous hand pose retargeting and interaction modeling by combining multi-source data and enforcing temporal and contact constraints.

Contribution

The paper presents a unified approach that integrates human hand and object data with differential loss and contact maps to improve retargeting precision and interaction realism.

Findings

Significant improvement in pose accuracy.

Enhanced naturalness and diversity of generated poses.

Robust modeling of hand-object interactions.

Abstract

Despite advances in hand-object interaction modeling, generating realistic dexterous manipulation data for robotic hands remains a challenge. Retargeting methods often suffer from low accuracy and fail to account for hand-object interactions, leading to artifacts like interpenetration. Generative methods, lacking human hand priors, produce limited and unnatural poses. We propose a data transformation pipeline that combines human hand and object data from multiple sources for high-precision retargeting. Our approach uses a differential loss constraint to ensure temporal consistency and generates contact maps to refine hand-object interactions. Experiments show our method significantly improves pose accuracy, naturalness, and diversity, providing a robust solution for hand-object interaction modeling.