Adaptive Volumetric Mechanical Property Fields Invariant to Resolution
2026-06-16 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine LearningRobotics
AI summaryⓘ
The authors propose AdaVoMP, a method that predicts detailed mechanical properties like stiffness, stretchiness, and density for 3D objects. Their approach uses a new smart voxel system called SAV, which works better and uses less memory than previous methods. Unlike earlier fixed-grid methods, AdaVoMP adaptively creates a unique material map for each object using a transformer model, allowing for much higher detail. Tests show it predicts properties more accurately and efficiently, enabling more realistic physics in simulations with complex 3D models.
Young's modulusPoisson's ratiodensity3D voxeltransformer modelautoregressive modelmaterial propertiesphysics simulationdeformable simulation
Authors
Rishit Dagli, Donglai Xiang, Vismay Modi, Xuning Yang, Gavriel State, David I. W. Levin, Maria Shugrina
Abstract
Accurate mechanical properties (or materials) Young's modulus ($E$), Poisson's ratio ($ν$) and density ($ρ$) are essential for reliable physics simulation of digital worlds, but most 3D assets lack this information. We propose AdaVoMP, a method for predicting accurate dense spatially-varying ($E$, $ν$, $ρ$) for input 3D objects across representations, improving the resolution, accuracy, and memory efficiency over the state-of-the-art. The foundation of our technique is a sparse and adaptive voxel structure SAV that efficiently represents both the input 3D shape and the material field output. We replace the fixed-voxel model of the most accurate prior method, VoMP, with a novel sparse transformer encoder-decoder model that learns to generate a unique SAV autoregressively for every input shape to represent its materials, achieving a resolution $16^3\times$ higher than prior art. Experiments show that AdaVoMP estimates more accurate volumetric properties, even with lesser test-time compute than all prior art. This allows us to convert high-resolution complex 3D objects into simulation-ready assets, resulting in realistic deformable simulations.