Subspace Neural Physics: Fast Data-Driven Interactive Simulation

preview_player
Показать описание

  1. Ubisoft La Forge
Рекомендации по теме
Subspace Neural Physics: 0:04:50

Subspace Neural Physics: Fast Data-Driven Interactive Simulation

Ubisoft's AI Learns 0:05:10

Ubisoft's AI Learns To Compute Game Physics In Microseconds! ⚛️

Data-driven Subspace Enrichment 0:02:27

Data-driven Subspace Enrichment for Elastic Deformations with Collisions

Data-driven model discovery: 0:45:38

Data-driven model discovery: Targeted use of deep neural networks for physics and engineering

Francesco Paesani: 'Data-driven 0:56:42

Francesco Paesani: 'Data-driven models for predictive molecular simulations'

Kevin Carlberg: AI 0:59:31

Kevin Carlberg: AI for Computational Physics: Toward real-time high-fidelity simulation

Youngsoo Choi: Efficient 0:58:49

Youngsoo Choi: Efficient physics-constrained data-driven physical simulations

Benjamin Peherstorfer - 1:00:37

Benjamin Peherstorfer - Physics-based machine learning for quickly simulating transport-dominated...

DDPS | libROM: 1:07:31

DDPS | libROM: Library for physics-constrained data-driven physical simulations | Youngsoo Choi

Simulating the Universe 0:23:31

Simulating the Universe with Machine Learning | AI & Physics | Shirley Ho

FastData@30012015PM 2:48:52

FastData@30012015PM

When it comes 0:05:40

When it comes to love, neural networks have a long way to go

DDPS | Efficient 0:05:23

DDPS | Efficient nonlinear manifold reduced order model

Physics-based domain adaptation 0:37:40

Physics-based domain adaptation for dynamical systems forecasting

Interpretable and Structure-Preserving 1:03:47

Interpretable and Structure-Preserving Data-Driven Methods for Physical Simulations - Youngsoo Choi

Using AI to 0:35:50

Using AI to Enable Rapid Simulation of Extreme-Scale Physics Models (Facebook)

[DDPS talk] #libROM: 1:07:31

[DDPS talk] #libROM: Library for physics-constrained data-driven physical simulations | #ROM #ML

Physics-Constrained NNs with 0:25:25

Physics-Constrained NNs with Minimax Architecture for Predicting Microstructure Evolution

Runhai Ouyang: The 0:23:36

Runhai Ouyang: The Data-driven Method SISSO: Concept, Challenges, and Applications

Kernel Analog Forecasting 0:38:56

Kernel Analog Forecasting for Multiscale Problems

Methods of Accelerating 2:24:56

Methods of Accelerating Physical Simulation||Ways to Design Intelligent Robotic System||June 4, 2021

Deep Learning to 0:26:56

Deep Learning to Discover Coordinates for Dynamics: Autoencoders & Physics Informed Machine Lear...

Kevin Carlberg - 1:10:09

Kevin Carlberg - AI for Computational Physics: Toward real-time high-fidelity simulation

The Diversity of 0:10:34

The Diversity of Neural Networks

Комментарии
Автор

This is probably one of the single best use cases for machine learning that I've seen so far. Yeah, you'll have to train the animations used in each game with the types of clothing physics you want for each different set of soft body clothes, and you'll have to train flags and things based on where you want them, etc, but once all the training is done, the final game can be just as good looking while being VASTLY more performant. I can't wait to see what this brings for the next generation of Assassin's Creed games. xD

KentHambrock
Автор

loving these applications of machine learning to game dev. been trying to do this myself, and I think you guys have the right approach

MsJeffreyF
Автор

Amazing work! I just finished reading your paper. Would you be able to share the source code with us?

alexandrepv
Автор

Using ImGui but not sharing any source code back, good job lol... proud of you... slow clap

kungfooman
Автор

Very cool! Shame the training time takes so long, but that's the cost of runtime overhead removal I guess.

spartanjackwar
Автор

you have killed nvidia's physx, did you get a job offer from them?

selfawaredevices
Автор

Add this in Blender and we're game, otherwise its just nice to look at and nothing more

anonymousd