Emerging Applications: Examples of Model-Based Computing

Model-based Computing Overview >
Interactive Realism for Learning and Entertainment >
Personal Media Management, Creation, Enhancement >
 
 
Model-based Computing Overview
 
The performance of future Tera-scale computers, based on 10s to 100s of integrated processor cores, will empower technology with more human-like capabilities. Our future devices will be able to comprehend data better and use this knowledge to act on our behalf. To do so, computers must have the ability to "think" in terms of models - digital "concepts" of the people, places, and things found within streams of bits and bytes.
By forming ever better models of what a person looks like in a photo, how virtual objects and characters should move, which sounds define speech, or even what data makes up a financial portfolio, future computers will be able to do much more than today. They will be able to recognize these models within rich data such as photos or streaming video. They will be able to mine your system or the internet for similar models. They will even be able to generate new models - synthesizing new virtual characters or testing out financial strategies for you. We call these capabilities RMS, for Recognition, Mining and Synthesis.
Our applications researchers are prototyping these new capabilities as software "workloads" to understand how to optimize them for Tera-scale architectures, guide future architectural designs, and demonstrate what you will be able to do with technology in 5 to 10 years.
Workload analysis
 
 
Interactive Realism for Learning and Entertainment back to top
 
 
Imagine watching yourself and your friends on the screen as life-like characters in your favorite movie or video game. You can pick-up and hold objects in the virtual world, push the villain out of the way, and reach out to grab the trophy as you proceed to the next level.
 
This video demonstrates how a Tera-scale computer could analyze images from multiple cameras in the home to capture body motions in 3D without any controller, special clothing, or blue screen in the background. The virtual character that mirrors the person's movements has been rendered using ray tracing techniques to display the scene more realistically - note the multiple reflections and shadows in the background. The ray tracing engine calculates the paths of individual light rays realistically using the laws of physics. This scene took many hours to generate on a powerful server, but future servers and PCs with Tera-scale processors will be able to do this in real time.
 
This video shows an example of physical modeling. This research, done in collaboration with Stanford University, models the movements of facial muscles underneath the skin to generate more realistic animations of human expressions. See more of this work at: graphics.stanford.edu/~sifakis/
 
 
Personal Media Management and Enhancement back to top
 
 
In the future you will be able to quickly access personal photos or videos by specifying the persons or objects in the images you want to find. Your Tera-scale computer will search for you, recognizing the content of the images automatically without requiring manual tagging or labeling of the content. Your PC will also be able to improve the quality of your videos.
 
Organizing and searching through large collections of personal images is becoming a significant challenge. In this video we demonstrate a Personal Image Retrieval Organizer which uses machine vision techniques to quickly search through more than 10,000 personal photos.
 
These videos show how a computer can enhance low-quality video (from a cell phone or the web). The application enhances the current video frame by using data from past and upcoming frames to reduce noise, remove camera shake and synthesize a better image. Future computers will be able to do this in real time.
 
 
Useful Links
Research@Intel Blog
News and Events
Tools and Software
Worldwide Network of Labs
University Research Grant Programs
 
 

 
Back to Top