Computing company Qualcomm Technologies has released a family of system-on-chips (SoCs) built for devices incorporating vision technology.
The Vision Intelligence Platform is designed for industrial and consumer smart security cameras, sports cameras, wearable devices, virtual reality, robotics, and smart displays, among other applications.
The QCS605 and QCS603 SoCs integrate an image signal processor (ISP), Qualcomm’s Artificial Intelligence (AI) Engine, an Arm-based multicore CPU, vector processor, and GPU. The platform also includes advanced camera processing software, machine learning and computer vision software development kits.
Qualcomm’s processors are typically found in mobile devices, but the company has recently released its Snapdragon 820 board, which will have availability and support until 2025 specifically to cater for the industrial market.
The new platform targets embedded vision IoT applications, which is a growing area both inside and outside the factory.
Joseph Bousaba, vice president, product management at Qualcomm Technologies, commented: ‘The Qualcomm Vision Intelligence Platform is the culmination of years of advanced research and development that brings together breakthrough advancements in camera, on-device AI and heterogeneous computing. The platform is a premier launch pad for manufacturers and developers to create a new world of intelligent IoT devices.’
The Vision Intelligence Platform integrates the Qualcomm AI Engine, which is comprised of several hardware and software components integrated to accelerate on-device AI. The AI Engine includes the Qualcomm Snapdragon Neural Processing Engine (NPE) software framework with analysis, optimisation, and debugging tools for developments using Tensorflow, Caffe and Caffe2 frameworks, Open Neural Network Exchange interchange format, as well as Android Neural Networks API and Qualcomm Hexagon neural network library – all of which are designed to allow developers and OEMs to port trained networks into the platform.
The new platform delivers up to 2.1 trillion operations per second of compute performance for deep neural network inferences. It supports up to 4K video resolution at 60 frames per second, or 5.7K at 30fps, as well as multiple concurrent video streams at lower resolutions.
The platform integrates a dual 14-bit Qualcomm Spectra 270 ISP, supporting dual 16 megapixel sensors. It also includes advanced vision processing capabilities necessary for IoT segments such as staggered HDR, electronic image stabilisation, de-warp, de-noise, chromatic aberration correction, and motion compensated temporal filters in hardware.
