Research map / 3D vision / 3D generation
3D-IntPhys: Towards More Generalized 3D-grounded Visual Intuitive Physics under Challenging Scenes
Haotian Xue, Antonio Torralba, Joshua B. Tenenbaum, Daniel LK Yamins, Yunzhu Li, Hsiao-Yu Tung
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 |
|---|---|---|---|
| use explicit 3D point-based perception instead of NeRF-dy | 3D-IntPhys · FluidPour InD | MSE: 202.95 -> 111.66 | Improved |
| use explicit 3D perception instead of NeRF-dy | 3D-IntPhys · FluidCubeShake OoD | MSE: 1585.97 -> 81.38 | Improved |
| use explicit 3D perception instead of NeRF-dy | 3D-IntPhys · GranularPush OoD | MSE: 1020.0 -> 646.85 | Improved |
| use 3D-IntPhys instead of autoencoder baseline | 3D-IntPhys · FluidCubeShake InD | SSIM: 0.71 -> 0.94 | Improved |
| use 3D-IntPhys instead of autoencoder baseline | 3D-IntPhys · GranularPush OoD | SSIM: 0.62 -> 0.69 | Improved |
| evaluate in extrapolate (OoD) versus in-distribution scenes | NeRF-dy · FluidCubeShake | SSIM: 0.73 -> 0.65 | Hurt |
| evaluate in extrapolate (OoD) versus in-distribution scenes | 3D-IntPhys · FluidCubeShake | SSIM: 0.94 -> 0.93 | No clear effect |
| evaluate in extrapolate (OoD) versus in-distribution scenes | autoencoder · GranularPush | MSE: 562.06 -> 1537.2 | Hurt |
| add spacing loss to dynamics model training (ablation) | 3D-IntPhys dynamics · GranularPush | Chamfer distance | Improved |
| use explicit 3D dynamics instead of NeRF-dy latent dynamics | 3D-IntPhys · long-horizon rollout, all scenarios | Chamfer distance | Improved |
| widen pusher by factor of 2 and 5 at test time | 3D-IntPhys dynamics · GranularPush | qualitative pushing prediction | No clear effect |
Closest papers by meaning
- Latent Intuitive Physics: Learning to Transfer Hidden Physics from A 3D Video (ICLR 2024 · similarity 0.91)
- FreeGave: 3D Physics Learning from Dynamic Videos by Gaussian Velocity (CVPR 2025 · similarity 0.89)
- Learning 3D Particle-based Simulators from RGB-D Videos (ICLR 2024 · similarity 0.89)
- DreamPhysics: Learning Physics-Based 3D Dynamics with Video Diffusion Priors (AAAI 2025 · similarity 0.88)
- Beyond Static Vision: Scene Dynamic Field Unlocks Intuitive Physics Understanding in Multi-modal Large Language Models (ICLR 2026 · similarity 0.87)
- OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation (ICLR 2025 · similarity 0.87)
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.
