PhysX-3D also proposes PhysXGen, a feed-forward framework for physics-grounded image-to-3D asset generation. PhysXGen employs a dual-branch architecture to explicitly model the latent correlations between 3D structures and physical properties, producing 3D assets with plausible physical predictions while preserving the native geometry quality. This framework enables the generation of 3D assets that can be used in various applications, including simulation and embodied AI, where physical properties are crucial.
PhysX-3D has been extensively evaluated, and the results demonstrate its superior performance and promising generalization capability. The framework has been shown to produce high-quality 3D assets that can be used in various applications. Additionally, PhysX-3D provides a scalable human-in-the-loop annotation pipeline based on vision-language models, enabling efficient creation of physics-first assets from raw 3D assets. This pipeline facilitates the development of 3D generative models that can produce physically plausible assets, paving the way for future research in generative physical AI.
