SAN FRANCISCO – March 7, 2023 – In the world of machine learning, it’s notoriously difficult to run models on a variety of cell phones. That’s why Fyusion, a visual intelligence company owned by Cox Automotive, has delivered a powerful new open-source library that makes it easy and efficient to run machine learning models on almost any mobile device. Now available on GitHub, this lightweight yet powerful software library called FyuseNet can use a variety of machine learning models and write incredibly fast implementations for the developer’s platform of choice – all while maintaining a lightweight application.
“At Fyusion, we live at the intersection of high-tech research and pragmatic engineering, bringing big ideas to life in ways that can really benefit those around us,” said Stephen Miller, co-founder and chief product officer of Fyusion. “We know how powerful FyuseNet is, having used it in our own work for years, and we believe this can be the one beautiful library that everyone uses to accelerate advances in machine learning.”
With stark differences between operating systems, as well as fragmentation across older and newer generations of phones, building machine learning models for mobile phones has been notoriously inefficient. This has led to innovation being mostly reserved for the $1,000+ flagship phones, which are out of reach for many people.
However, the team at Fyusion is on a mission to democratize machine learning and ensure these technologies are more accessible to more people. For this reason, it was important for them to open access to FyuseNet, since its easy implementation makes it compatible with most devices – regardless of manufacturer or specific processor details – even when no network connection is available. FyuseNet was developed in 2016 for internal use and has been iterated within Fyusion for years before making its first appearance on GitHub this year.
The technical data:
FyuseNet is an OpenGL(ES)-based library that enables neural network inference to run on GPUs that support OpenGL and OpenGL(ES). Written in portable C++, the library runs on a variety of desktop, edge, and mobile platforms. FyuseNet is made freely available under the MIT license.
In order to deliver the performance required for real-time execution, the developers have built a number of optimizations into the library, the most important being the ability to merge some operations into a single layer/shader. A second trick FyuseNet employs is to take advantage of the GPU’s raster operations processors. This has the additional advantage that some arithmetic operations are essentially free, since the execution time inside the shader is not changed. The runtime has no significant external dependency other than OpenGL, giving users clean, simple, and extremely portable code.
“When we founded Fyusion, one of our goals was to leverage the immense technical expertise of our team to solve real-world problems faced by everyday people,” added Radu Rusu, co-founder and CEO of Fyusion. “And as innovations in machine learning happen fast, we believe making FyuseNet freely available will help us build a broader community of developers who can contribute together as discoveries are made, and ensure even more people in the coming ones.” be able to benefit from it for years to come.”
Headquartered in San Francisco, Fyusion employs some of the world’s leading machine learning and AI researchers and developers, helping customers view and analyze millions of 3D automotive images each year. Acquired in 2020 by Cox Automotive, the world’s largest automotive services and technology provider, Fyusion remains at the forefront of computer vision development, leveraging software, imaging technology, sensors and services to extract actionable insights from visual data.
As part of Cox Automotive, Fyusion has delivered the only AI technology for automotive wholesale that uses 3D imaging to view vehicle damage from every angle. Fyusion’s technology is used in both mobile and stationary imaging solutions to capture millions of used vehicles each year for the world’s largest wholesale marketplace, Manheim. Together, Fyusion and Manheim use the same technology in FyuseNet to provide the automotive industry’s most advanced automated damage detection solution for used vehicles.
About Fusion
Fyusion is a computer vision company that develops innovative solutions for vehicle inspection, vehicle imaging and related applications. With over 150 patents and patents pending, Fyusion is a pioneer in transforming visual data into actionable information. Based in San Francisco, our company includes some of the world’s leading researchers and developers in machine learning and AI. Founded in 2014, Fyusion is a brand of Cox Automotive™. Visit www.fyusion.com for more information.
About Cox Automotive
Cox Automotive is the world’s largest automotive services and technology provider. Powered by the broadest range of first-party data, fed by 2.3 billion online interactions per year, Cox Automotive develops leading-edge solutions for car buyers, automakers, dealers, retailers, lenders and fleet owners. The company has more than 25,000 employees on five continents and a family of trusted brands that includes Autotrader®, Dealertrack®, Kelley Blue Book®, Manheim®, NextGear Capital™ and vAuto®. Cox Automotive is a subsidiary of Cox Enterprises Inc., a $22 billion, privately held Atlanta-based company. Visit coxautoinc.com or connect with @CoxAutomotive on Twitter, CoxAutoInc on Facebook, or Cox-Automotive-Inc on LinkedIn.
Media contacts:
Julie Zorn Shipp
Senior Manager, Public Relations, Cox Automotive
404-558-7837
[email protected]
Kelsey Guy
Sr. Marketing and Communications Manager
fusion
(214) 538-2301
[email protected]
