Author: Apagom AG

Physics Forests Demo on Linux

Physics Forests Demo has been released on Linux 64-bit, compiled on Ubuntu 16. Linked to the Download Page    

Fluid Interactions with Animating Convex Decomposition

The approximate decomposition of the animated mesh into non-rigid convex parts for fast collision detection (physics, rendering). Animations were pre-processed (trained) in 30s – 4min. The approximation error can be chosen afterwards interactively (0-2s). Coarse approximation is suitable for real-time applications, finer for offline simulations or ray tracing. The method is robust to topologically incoherent…
Read more

Physics Forests in VR

Implemented using Oculus SDK and CUDA, rendered by OpenGL shader. Real-time simulation with above 2M particles in 60~90 fps. Obstacles, swirls and tornado effects are controlled by the Oculus touch controllers.

Artistic Controls: Turbulence

Create various water scenes with controllable turbulence. The demo runs 30FPS with 5.1M particles on a single NVidia Pascal TitanX GPU.

Endless Simulation

Physics Forests always provides the stable behavior of fluids in an arbitrary long video. Simulation never diverges with extreme user interactions or external forces.

Various Materials

Physics Forests can simulation a variety of liquids with different material properties such as viscosity, adhesion, surface tension, friction and other external forces.

Real-time Simulation using Machine Learning

Physics Forests is the fastest and stable physics-based fluid simulation engine based on Machine Learning, capable of simulating up to 10M in real-time.

SIGGRAPH 2017 REALTIME LIVE!

“Physics Forests: Real-time Fluid Simulation using Machine Learning” has been accepted to SIGGRAPH 2017 Real-time live! SIGGRAPH Real-time Live! Demo Presentation Videos    

Physics Forests

The Fastest Real-time Fluid Engine using Machine-Learning https://vimeo.com/209397181

Real-time Simulation using Machine-Learning

Physics ForestsĀ is the fastest stable physics-based fluid simulation engine based on Machine Learning, capable of simulating up to 10M particles including surface generation, rigid-body dynamics, rendering and user-interaction in real-time. This video was recorded on a standard PC with a single Nvidia Pascal Titan X (Cuda).