Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces ODeform, an innovative method that leverages Neural Ordinary Differential Equations to model continuous 4D motion for shape deformation in 3D space. By transforming 3D point clouds and physical properties into a unified latent space, ODeform enables real-time simulation of object deformations without relying on discrete time steps, thus enhancing computational efficiency. The method outperforms baseline approaches in motion prediction accuracy and successfully generalizes to novel 3D shapes, demonstrating its robustness in practical applications.
ODeform achieves real-time continuous shape deformation modeling, outperforming traditional methods by eliminating discrete time steps.
Modeling continuous object deformation is important for many computer vision and robotics tasks, such as manipulation and simulation. Existing approaches rely on learning-based methods or physics simulators to model shape deformations. However, these approaches either use discrete time steps or are too computationally intensive for real-time applications. We present ODeform, a novel extension of Neural Ordinary Differential Equations to continuous 4D dynamics of deformable objects in 3D space. Our method transforms 3D point clouds and physical conditions (like material properties) into a unified latent space. By solving the resulting ordinary differential equations over time, we model deformations as continuous flows within this learned embedding, eliminating the need for discrete time steps while maintaining computational efficiency. We evaluate our approach on unseen physical parameter configurations, showing improved motion prediction accuracy over baseline methods. Our experiments further demonstrate a successful transfer to real 3D captured objects with novel shapes, along with effective interpolation and extrapolation of the learned dynamics. Our code and data will be made publicly available.