Computer software mimics the brain's vision
Left: A roadrunner, the New Mexico state bird, used as input to the NeuralViz program. Right: NeuralViz model of neural activity across primary visual cortex. NeuralViz uses biological learning rules to build brain-like hierarchical representations of natural imagery that can support accurate and robust extraction of objects from images.
Contact
- David Seigel
- Technology Transfer
- (505) 665-2743
Seeing with the brain: developing computer software to mimic the brain's vision
Applications:
NeuralViz can be used for robust object recognition in imagery. Potential markets include automated monitoring of video and still imagery for:
- content-based search
- security, safety, and inspection applications
- aerial / satellite remote sensing for defense
- intelligence and civilian applications
- and autonomous robots.
Benefits:
The speed at which the algorithms can recognize objects as well as learn new objects (to add to the library of recognized objects) is unlike any other tool available.
Intellectual property status:
NeuralViz is protected by international copyright and LANL is seeking patent protection on the algorithms related to the technology also.
Licensing status:
Available for licensing
Summary:
Scientists at LANL are exploring how the brain sees. Their goal is to develop computer vision software with human-level speed, accuracy and robustness.
NeuralViz is an evolving software technology developed by the Synthetic Visual Cognition Team at Los Alamos National Laboratory (LANL) to demonstrate full-scale, real-time models of visual cortex, the part of the human brain responsible for vision. The code runs on Roadrunner, LANL’s 1,000-teraflop (“petaflop”) super-computer.
Small mammals are also capable of excellent visual acuity and object recognition with brains orders of magnitude smaller than humans. Thanks to the revolution in cluster-on-a-chip computing driven by the gaming/entertainment industry, NeuralViz can also run visual cortex models of these animals using commercial off-the-shelf workstations. For example, a graphical processing unit (GPU) for a standard workstation provides 1 teraflop of computing power for less than $500 and has the processing power of a mouse’s visual cortex. NeuralViz thus has the potential to enable widespread use of fast, accurate, robust computer vision systems for a wide range of real-world applications.
Development stage:
Team publications are available at http://synthetic-cognition.lanl.gov. As computing technology increases in computing power and decreases in size, weight and power requirements, the applications for NeuralViz will continue to grow.
Related Articles
All Stories
Bees replacing dogs as detectors of volatile organic compounds
August, 6 2012 - Researchers at LANL began working with domestic honeybees in 2006 to detect explosive materials and narcotics.
Increased durability and lifetime of fuel cell products
April, 3 2012 - A revolutionary method of building a membrane electrode assembly (MEA) for PEM fuel cells allows LANL to offer significant cost reductions and performance enhancements necessary for mainstream applications.
Digital content tool analyzes "gist" of text-based documents
April, 3 2012 - LANL’s Digital Knowledge Discovery Team has created a suite of digital-content-analysis tools to gather, reduce, annotate, organize, synthesize, and visualize digital content.
Reinvention of laboratory apparatus cuts costs
April, 3 2012 - New apparatus that is designed to be inexpensive, reconfigurable, and adaptable to both surface and solution electrochemical experiments.
Solving the energy overusage problem
April, 3 2012 - Energy use and related thermal management issues are serious business concerns, but with EnergyFit LANL can significantly reduce heat output and maintain consistent electrical cost.
Hazardous odorless gases detected by LANL's IntelliCLAD
April, 3 2012 - This new class of smart coatings are non-odorous but release an odor in the presence of selected gases of concern, reducing contamination and alerting people when danger arises.
Programers advance with Kip Ray Tracer
April, 3 2012 - Kip is a high-performance ray tracing library that provides rapid rendering and compact code while using very little memory so large numbers of shapes can be used simultaneously.
KIVA's modeling code increases engine efficiency and fuel economy
April, 3 2012 - Through an advanced computational modeling code that accurately simulates the in-cylinder processes of engines, LANL scientists have addressed many costly issues.
Deep-space communication improved with electromagnetic radiation antenna
April, 3 2012 - New antenna has potential applications in RADAR, secure communications, ultra-long-range communications, oncology and astrophysics.
Faster than the speed of sight: High-speed camera provides high sensitivity, frame rate
April, 3 2012 - High-speed cameras becomes less effective at high speeds; MOXIE fixes that problem by simultaneously providing both the highest photographic and physical speeds.
Video
Small Reactor for Deep Space Exploration
This is the first demonstration of a space nuclear reactor system to produce electricity in the United States since 1965, and an experiment demonstrate...
