JHU unveils PowerSim: differentiable physics that simulates and re-renders captured 3D scenes
anand_bhattad · x · 2026-10-01
PowerSim, the first representative paper from the new 3P Vision Group at Johns Hopkins (Trong-Tung Nguyen, Anand Bhattad), adds physically grounded differentiable dynamics to PowerFoam's power-diagram-based 3D representation by directly coupling pre-trained scenes with the Material Point Method (MPM) — simulated motion drives geometry and appearance with no auxiliary representation.
Capabilities:
- Physically interact with objects in captured scenes
- Recover spatially varying material fields
- Composite objects from independently captured scenes into one simulation-ready scene
- Ray-traced reflections that stay consistent as objects deform
Benchmarks: under extreme stretching, PAC-NeRF loses texture and shows cut-like boundaries while PhysGaussian blurs; PowerSim's explicit volume cells preserve sharper detail, rotating dipole planes with the material so appearance follows deformation correctly. Project page: power-sim.github.io; code TBD.
More from Multimodal
- Failed MiniMax H3 VAE detail experiment yields a useful 2X detail VAE and workflow — NoMouse9610 · 2026-10-01
- ThinkV2V: Reasoning-Driven Video Editing — a 5B Model Beats 10B Baselines on Complex Instructions — Donghao Zhou · 2026-10-01
- Fudan's IDSpect Decomposes Chinese Characters into Radicals for Fine-Grained Text-to-Image RL Rewards — Fudan-University · 2026-10-01
- Imagine3D-LLM teaches MLLMs to imagine a 3D scene before answering — kaist-ai · 2026-10-01
- 7000 Midjourney Images + Suno Track + Opus 5.5: 2-Hour AI Creation Cost Just $6.80 — ciguleva · 2026-10-01
- ST-AudioLM From Sony AI and POSTECH Tracks Moving Sound Sources With 40 Trajectory Tokens — mittu1204 · 2026-10-01