Research map / 3D vision / 3D generation
TRACE: Learning 3D Gaussian Physical Dynamics from Multi-view Videos
Jinxi Li, Ziyang Song, Bo Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add auxiliary deformation field f_defo (ablation) | TRACE · Dynamic Multipart | PSNR: 19.206 -> 33.481 | Improved |
| derive future motion from physics parameters at t' rather than querying for each future time (ablation) | TRACE · Dynamic Multipart | PSNR: 29.602 -> 33.481 | Improved |
| use equivalent velocity/acceleration parametrization (ablation) | TRACE · Dynamic Multipart | PSNR: 29.986 -> 33.481 | Improved |
| use training time difference 2 delta t instead of delta t (ablation) | TRACE · Dynamic Multipart | PSNR: 32.852 -> 33.481 | Improved |
| use 2 delta t instead of 3 delta t (ablation) | TRACE · Dynamic Object | PSNR: 29.787 -> 31.511 | Improved |
| use 2nd-order instead of 1st-order dynamics (ablation) | TRACE · Dynamic Object | PSNR: 28.522 -> 31.511 | Improved |
| use 2nd-order instead of 3rd-order dynamics (ablation) | TRACE · Dynamic Indoor Scene | PSNR: 34.086 -> 34.824 | Improved |
| learn hard physics parameters instead of PINN soft constraints (vs DefGS_nvfi) | TRACE · Dynamic Multipart | PSNR: 28.455 -> 33.481 | Improved |
| use 4DGS as auxiliary deformation field | TRACE · NVIDIA Dynamic Scene | PSNR: 29.341 -> 22.554 | Hurt |
| replace baseline deformable 3DGS with TRACE dynamics in continual learning | DefGS_nvfi · ParticleNeRF | PSNR: 26.181 -> 27.544 | Improved |
| cluster learned physical parameters for motion segmentation vs NVFi | NVFi · Dynamic Indoor Scene | AP: 91.21 -> 95.82 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- NVFi — NVFi: Neural Velocity Fields for 3D Physics Learning from Dynamic Videos (NeurIPS 2023) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FreeGave: 3D Physics Learning from Dynamic Videos by Gaussian Velocity (CVPR 2025 · similarity 0.92)
- DreamPhysics: Learning Physics-Based 3D Dynamics with Video Diffusion Priors (AAAI 2025 · similarity 0.89)
- Learning 3D Particle-based Simulators from RGB-D Videos (ICLR 2024 · similarity 0.89)
- Motion4D: Learning 3D-Consistent Motion and Semantics for 4D Scene Understanding (NeurIPS 2025 · similarity 0.88)
- OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation (ICLR 2025 · similarity 0.88)
- Motion Decoupled 3D Gaussian Splatting for Dynamic Object Representation (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
