Research map / 3D vision / Face and shape
RISP: Rendering-Invariant State Predictor with Differentiable Simulation and Rendering for Cross-Domain Parameter Estimation
Pingchuan Ma, Tao Du, Joshua B. Tenenbaum, Wojciech Matusik, Chuang Gan
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 the rendering-gradient regularizer to domain-randomized state prediction (ablation) | RISP · quadrotor state estimation | state estimation error: 0.3114 ± 0.3191 -> 0.1505 ± 0.1163 | Improved |
| add the rendering-gradient regularizer to domain-randomized state prediction (ablation) | RISP · cube state estimation | state estimation error: 0.2805 ± 0.3199 -> 0.1642 ± 0.1887 | Improved |
| add the rendering-gradient regularizer to domain-randomized state prediction (ablation) | RISP · hand state estimation | state estimation error: 0.1155 ± 0.0505 -> 0.0974 ± 0.0255 | Improved |
| add the rendering-gradient regularizer to domain-randomized state prediction (ablation) | RISP · rod state estimation | state estimation error: 0.0201 ± 0.0087 -> 0.0194 ± 0.0048 | Improved |
| add the rendering-gradient regularizer (ablation) | RISP · quadrotor system identification | mass estimation error: 6.07e-2±4.34e-2 -> 1.18e-2±1.93e-2 | Improved |
| add the rendering-gradient regularizer (ablation) | RISP · cube system identification | stiffness estimation error: 1.16e-1±6.20e-2 -> 6.76e-3±7.23e-3 | Hurt |
| add the rendering-gradient regularizer (ablation) | RISP · hand system identification | joint stiffness estimation error: 4.85e-2±2.16e-2 -> 3.96e-2±2.73e-2 | Improved |
| add the rendering-gradient regularizer (ablation) | RISP · rod system identification | Young's modulus estimation error: 8.78e4±1.52e5 -> 9.31e1±4.21e1 | Improved |
| add the rendering-gradient regularizer (ablation) | RISP · quadrotor imitation learning | state discrepancy: 25.07 ± 5.76 -> 2.63 ± 1.86 | Improved |
| add the rendering-gradient regularizer (ablation) | RISP · hand imitation learning | state discrepancy: 7.12 ± 0.71 -> 1.52 ± 0.18 | Improved |
| add the rendering-gradient regularizer (ablation) | RISP · rod imitation learning | state discrepancy: 0.83 ± 0.35 -> 1.05 ± 0.27 | Hurt |
| train with rendering-gradient regularization rather than domain randomization alone (ablation) | RISP · quadrotor state estimation with varying numbers of rendering configurations | state estimation loss | Improved |
Closest papers by meaning
- Learning Physics From Video: Unsupervised Physical Parameter Estimation for Continuous Dynamical Systems (CVPR 2025 · similarity 0.85)
- Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representations (ICML 2025 · similarity 0.84)
- Accelerating Visual-Policy Learning through Parallel Differentiable Simulation (NeurIPS 2025 · similarity 0.84)
- SurMo: Surface-based 4D Motion Modeling for Dynamic Human Rendering (CVPR 2024 · similarity 0.84)
- DynaRend: Learning 3D Dynamics via Masked Future Rendering for Robotic Manipulation (NeurIPS 2025 · similarity 0.84)
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.83)
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.
