June 29, 2016 -
Sighthound announced that it has released an advanced facial recognition software for mobile handheld platforms.
The company will be demonstrating cellphone and tablet based implementations at this week’s CVPR conference in Las Vegas, Nevada.
Sighthound focuses on developing deep networks that run in real time and require less training data than its industry competitors.
According to Sighthound, traditional deep networks are computationally expensive and therefore require cloud based server arrays or GPU accelerated hardware.
In contrast, Sighthound’s networks are designed to run on standard CPUs, and on handheld devices, without the algorithms compromising any performance or accuracy.
Sighthound says that its algorithms compete with or exceed the best academic and commercial systems currently available based on benchmarking tests on publicly available data sets.
At CVPR, Sighthound will demonstrate face detection, recognition and tracking software on iPhone 6 and iPad Pro.
The software — which is designed for OEMs and developers interested in incorporating individual identification and verification features in their products and applications — can perform real time recognition on live video feeds.
Additionally, the software is not restricted to still images or locked inside Apple applications, nor does it rely on a connection to cloud servers.
“Offering scalable computer vision software implementations from the cloud to embedded portable devices is essential to our value proposition,” said Stephen Neish, CEO of Sighthound. “Our mantra is to make computer vision easy to use — that now includes customers who want to deploy solutions on mobile devices.”
Sighthound’s facial detection, recognition and tracking software is the latest addition to the Sighthound Sentry suite of computer vision capabilities.
In addition, Sighthound offers a smart home and business surveillance system, Sighthound Video, which uses its facial detection software, and Sighthound Cloud, a set of computer vision APIs for developers.