FSAG: Enhancing Human-to-Dexterous-Hand Finger-Specific Affordance Grounding via Diffusion Models
By: Yifan Han , Pengfei Yi , Junyan Li and more
Dexterous grasp synthesis remains a central challenge: the high dimensionality and kinematic diversity of multi-fingered hands prevent direct transfer of algorithms developed for parallel-jaw grippers. Existing approaches typically depend on large, hardware-specific grasp datasets collected in simulation or through costly real-world trials, hindering scalability as new dexterous hand designs emerge. To this end, we propose a data-efficient framework, which is designed to bypass robot grasp data collection by exploiting the rich, object-centric semantic priors latent in pretrained generative diffusion models. Temporally aligned and fine-grained grasp affordances are extracted from raw human video demonstrations and fused with 3D scene geometry from depth images to infer semantically grounded contact targets. A kinematics-aware retargeting module then maps these affordance representations to diverse dexterous hands without per-hand retraining. The resulting system produces stable, functionally appropriate multi-contact grasps that remain reliably successful across common objects and tools, while exhibiting strong generalization across previously unseen object instances within a category, pose variations, and multiple hand embodiments. This work (i) introduces a semantic affordance extraction pipeline leveraging vision-language generative priors for dexterous grasping, (ii) demonstrates cross-hand generalization without constructing hardware-specific grasp datasets, and (iii) establishes that a single depth modality suffices for high-performance grasp synthesis when coupled with foundation-model semantics. Our results highlight a path toward scalable, hardware-agnostic dexterous manipulation driven by human demonstrations and pretrained generative models.
Similar Papers
Towards Affordance-Aware Robotic Dexterous Grasping with Human-like Priors
Robotics
Robot hands learn to grab things like humans.
Towards Affordance-Aware Robotic Dexterous Grasping with Human-like Priors
Robotics
Teaches robot hands to grab things like humans.
Towards Affordance-Aware Robotic Dexterous Grasping with Human-like Priors
Robotics
Robot hands learn to grab objects like humans.