Application-specific neural network allows video to be processed more like a human brain
HILLSBORO, OR, February 16, 2016 /24-7PressRelease/ — Today VEFXi Corporation announces a radical new approach to video processing using an application-specific neural network design, as opposed to the older approach using general processors and software. This new architecture can be applied to a wide range of real-time video processing applications up to 4K UHD. A tablet-based demonstration is now available using the patent-pending approach applied to converting live streaming 2D video and gaming to no-glasses 3D. Potential customers have already described the demo as “holographic-like” with 3D popout and depth “similar to custom handcrafted 3D” even when converting from 2D.
Radical New Approach
The chip design development has taken years to perfect. Instead of using a software/processor-based approach, VEFXi’s design works more like the human brain. There are no processors or instructions. The design is divided into pipelined stages to process video in real-time. Each stage has small memory elements which can remember data that has come before computation and previous computed data. Output from one stage feeds directly into the next stage. No external memory is required and maximum efficiency is achieved.
Unlike a processor that does one calculation at a time, VEFXi’s pipelined architecture makes thousands of calculations at a time but at 1/10th the rate. Because so much work is going on simultaneously the resulting computation can be comparable to a hundred or more processors. It requires far fewer operations for each computation compared to a processor making this approach dramatically more power efficient. Because of the slower rate the VEFXi microchip is easy to design with conventional ASIC synthesis design approaches, which are faster and easier than the complicated and expensive custom circuit-intensive approaches needed for a processor design. This results in a smaller die size, lower power, and lower cost.
VEFXi pushed this design approach even further on this generation by inventing their patent-pending NeuralBrain Depth Synthesis technology which is used to create the depth position of each pixel in the resulting 3D output. Instead of the conventional approach which requires an army of software programmers to develop algorithms, VEFXi applies logic which even more closely models the functions of the human brain neuron cells and associated synapses. This allows not only layering of algorithms but the ability to train or teach the logic to make better decisions. During training of this type, the hardware brain learns new algorithms.
OEMs can significantly improve user 3D experience for no-glasses 3D mobile phones, tablets, laptops, computers, and big screen TVs. The technology can also be used to improve both VR and AR products.
The VEFXi Technical Team
VEFXi’s CEO and founder, Craig Peterson, worked at Intel for 29 years where he led the development of various microprocessors, chipsets, and founded three new internal business startups including two supercomputer microchip programs with DARPA funding. Other VEFXi development team members included engineers who have worked on various high-speed video processing technologies at Northrup Grumman, FLIR, Tektronix, Nokia, Intel, and NVIDIA.
See for Yourself
See VEFXi’s demo at Mobile World Congress in Barcelona February 22-25, booth 8.0 K41.
VEFXi Corporation develops and sells embedded microchips that convert live 2D or 3D video to no-glasses 3D video. VEFXi was founded in December 2010 and has shipped products to more than 30 countries and is now demonstrating its new third generation converter technology.
Read the original story here: http://www.24-7pressrelease.com/press-release/radical-new-video-microchip-design-approach-419217.php
Press Release Distributed by 24-7 Press Release Newswire