GAPartManip: A Large-scale Part-centric Dataset for Material-Agnostic Articulated Object Manipulation

The article introduces GAPartManip, a large-scale part-centric dataset designed for material-agnostic articulated object manipulation, featuring photo-realistic material randomization and detailed annotations for actionable interaction poses. The dataset was evaluated by integrating it with state-of-the-art methods for depth estimation and interaction pose prediction, revealing significant performance improvements in both simulated and real-world environments. This resource is crucial for AI practitioners aiming to enhance the robustness and adaptability of manipulation systems in complex household scenarios.