Excellent responsiveness with low latency from sensor to inference control
You can develop applications without hardware knowledge.

Xilinx announced that it is expanding its vision-based machine learning applications with the Xilinx reVISION stack.

The Xilinx reVISION stack complements and expands the vision-based scope of the reconfigurable acceleration stack, which was launched last November and targets a variety of compute acceleration applications, including machine learning inference.

The reVISION stack enables software and systems engineers to quickly and easily develop intelligent vision-based systems, requiring little to no hardware design expertise. This makes it easier for engineers to combine machine learning, computer vision, sensor fusion, and connectivity.


In markets where rapid deployment of cutting-edge algorithms and sensors is essential, reVISION can support rapidly growing applications. Approximately two-thirds of applications involve vision-centric semiconductors. Applications as diverse as high-end consumer electronics, automotive, industrial, medical, aerospace, and defense will see the combination of machine learning and sensor fusion powering next-generation collaborative robots (cobots), sense-and-avoid drones, augmented reality, autonomous vehicles, automated surveillance, and medical diagnostics.

To reduce reliance on hardware experts, we launched the SDSoC development environment based on C, C++, and OpenCL. However, this still falls short of widespread adoption, deployment, and the complexity of machine learning. To address this, the reVISION stack includes a variety of development resources for platform, algorithm, and application development. This includes support for popular neural networks such as AlexNet, GoogLeNet, SqueezeNet, SSD, and FCN.

It also provides optimized library components for the CNN network layers required to build custom neural networks (DNNs/CNNs). These machine learning components are complemented by a diverse set of accelerated OpenCV functions for computer vision processing.

Xilinx supports industry-standard frameworks such as Caffe for machine learning and OpenVX for computer vision for application-level development. The reVISION stack also includes several types of sensors and development platforms from Xilinx and third parties.

“Xilinx is seeing tremendous interest in machine learning from the edge to the cloud, and our continued investments in our development stack will accelerate this adoption,” said Steve Glaser, senior vice president, Corporate Strategy Group, Xilinx. “Today, hundreds of embedded vision customers are leveraging Xilinx technology to achieve 10x or more performance and latency improvement. With the addition of reVISION, thousands more will now benefit.”