Market Context

 

Embedded Vision is one of the most exciting fields in technology today. Xilinx sees embedded vision as a key and pervasive megatrend that is shaping the future of the electronics industry.

 

Providing machines the ability to see, sense, and immediately respond to the world creates unique opportunities for system differentiation; however, this also creates challenges in how designers create next-generation architectures and bring them to market. Integrating disparate sub-systems including video and vision I/O with multiple image processing pipelines, and enabling these embedded-vision systems to perform vision-based analytics in real time is a complex task that requires tight coordination between hardware and software teams. To remain timely and relevant in the market, leading development teams are exploiting Xilinx’s devices in their next-generation systems to take advantage of the devices’ programmable hardware, software, and I/O capabilities.

 

Xilinx Solutions for Embedded Vision

 

Xilinx provides embedded vision developers with a suite of technologies that support both hardware and software design. Xilinx devices include FPGAs, SoCs and MPSoCs.

 

The Xilinx Vivado HLx design environment supports both hardware and platform developers developing the latest embedded-vision hardware. These tools include support for the industry’s latest high-bandwidth sensor interfaces. Xilinx SDx tools including SDSoC allows software and algorithm developers to develop in familiar Eclipse-based environments in familiar languages like C, C++ and OpenCL.

 

The Xilinx reVISION Stack builds upon the SDx concept to include support for OpenCV and machine learning inference, including support for the most popular neural networks such as AlexNet, GoogLeNet, SqueezeNet, SSD, and FCN as well as the functional elements required to build custom neural networks (CNNs/DNNs) while permitting design teams to leverage pre-defined and optimized CNN implementations for network layers. This is complemented by a broad set of acceleration-enabled OpenCV functions for computer vision processing.