Search papers, labs, and topics across Lattice.
This paper addresses the data scarcity bottleneck in training robust dexterous grasping models for cluttered scenes by introducing OmniDex, a benchmark of 2.6 million diverse scenes with 0.4 billion grasp ground truths generated via a scalable seed-and-filter strategy. The proposed OmniDex model overcomes grasp multimodality and precision errors by employing Soft Winner-Takes-All learning with physical constraints and physics-driven ranking, achieving state-of-the-art performance without post-optimization latency. This work significantly advances the state of embodied AI by enabling scalable, high-quality data generation and a more robust grasping model for complex real-world environments.
A massive 2.6M-scene benchmark and a novel generative model finally unlock robust dexterous grasping in cluttered environments, outperforming prior methods without post-optimization.
Dexterous grasping is the foundational primitive in embodied AI, demanding massive data to train robust models. As real-world data collection is expensive, simulation has become the mainstream paradigm. Yet, while cluttered scenes best reflect real-world applications, learning to grasp within them is bottlenecked by a critical scarcity of large-scale data. To resolve this, we curate high-quality 3D objects and supporting bases, proposing a scalable seed-and-filter strategy that bypasses sluggish scene-level optimization. This yields an unprecedented benchmark comprising over 2.6 million scenes and 0.4B scene-specific grasp ground truths, featuring diverse realistic layouts paired with rich semantic and geometric observations. Furthermore, we introduce the OmniDex model to overcome the grasp multimodality and last-millimeter precision errors plaguing current generative models. By coupling Soft Winner-Takes-All learning with human-inspired physical constraints during training, and utilizing physics-driven ranking, our approach achieves robust dexterous grasping without the latency of post-optimization. Experimental results show that OmniDex model achieves state-of-the-art performance and strong generalization across diverse scenes, views, and unseen objects.
