Physical Property
Research on physical property estimation focuses on accurately determining material characteristics from various data sources, including visual observations, acoustic signals, and robotic interactions. Current efforts leverage machine learning models, such as physics-informed neural networks, graph neural networks, and video diffusion models, to infer properties like mass, elasticity, friction, and material composition, often incorporating physics-based simulations for improved accuracy. This work is significant for advancing fields like robotics, computer vision, and materials science, enabling more realistic simulations, improved object manipulation, and more efficient material characterization.
Papers
In-Context Learning of Physical Properties: Few-Shot Adaptation to Out-of-Distribution Molecular Graphs
Grzegorz Kaszuba, Amirhossein D. Naghdi, Dario Massa, Stefanos Papanikolaou, Andrzej Jaszkiewicz, Piotr Sankowski
DreamPhysics: Learning Physical Properties of Dynamic 3D Gaussians with Video Diffusion Priors
Tianyu Huang, Haoze Zhang, Yihan Zeng, Zhilu Zhang, Hui Li, Wangmeng Zuo, Rynson W. H. Lau