Microchip neural network IP helps embed vision

  • May 20, 2020
  • imc

Arizona-based Microchip has revealed a software development kit and neural network IP for creating low-power FPGA smart embedded vision products.
The VectorBlox SDK and IP provide an easier way for software developers to programme a trained neural network without prior FPGA expertise.
With the rise of artificial intelligence (AI), machine learning and the IoT, applications are moving to the network edge where data are collected, requiring power-efficient products to deliver more computational performance in smaller, thermally constrained form factors.
Through its Smart Embedded Vision initiative, Microchip Technology says it is meeting the growing need for power-efficient inferencing in edge applications by making it easier for software developers to implement their algorithms in PolarFire field-programmable gate arrays (FPGAs).
The VectorBlox accelerator software development kit (SDK) helps developers take advantage of Microchip’s PolarFire FPGAs for creating low-power, flexible overlay-based neural network applications without learning an FPGA tool flow.
FPGAs are suitable for edge AI applications, such as inferencing in power-constrained compute environments, because they can perform more giga operations per second with greater power efficiency than a central or graphics processing unit, but they require special hardware design skills. The VectorBlox accelerator SDK is designed to enable developers to code in C and C++ and programme power-efficient neural networks without prior FPGA design experience.
The tool kit can execute models in TensorFlow and the Onnx open neural network exchange format, which offers wide framework interoperability. Onnx supports many frameworks such as Caffe2, MXNet, PyTorch and Matlab. The SDK is supported on Linux and Windows operating systems, and it includes a bit accurate simulator that provides the user the opportunity to validate the accuracy of the hardware while in the software environment.
The neural network IP included with the kit also supports the ability to load different network models at run time.
“For software developers to benefit from the power efficiencies of FPGAs, we need to remove the impediment of them having to learn new FPGA architectures and proprietary tool flows, while giving them the flexibility to port multi-framework and multi-network solutions,” said Bruce Weyer, vice president at Microchip. “Microchip’s VectorBlox accelerator SDK and neural network IP core will give both software and hardware developers a way to implement an extremely flexible overlay convolutional neural network architecture on PolarFire FPGAs, from which they can then more easily construct and implement their AI-enabled edge systems that have best-in-class form factors, thermals and power characteristics.”
For inferencing at the edge, PolarFire FPGAs are said to deliver up to 50 per cent lower total power than competing devices, while offering 25 per cent higher-capacity maths blocks that can deliver up to 1.5 tera operations per second. By using FPGAs, developers also have greater opportunities for customisation and differentiation through the devices’ inherent upgradability and ability to integrate functions on a single chip.
The PolarFire FPGA neural network IP is available in a range of sizes to match the performance, power and package size trade-offs for the application, enabling users to implement their designs in package sizes as small as 11 by 11mm.
The Smart Embedded Vision initiative was launched last July to provide hardware and software developers with tools, IP cores and boards for meeting the thermally constrained and small-form-factor requirements of edge applications. Because PolarFire FPGAs deliver lower power, users can eliminate the need for fans in their enclosures. The FPGAs also have more functional integration for a design.
For example, in applications such as a smart camera, the FPGAs can integrate the image signal pipeline, which includes the sensor interface, DDR controller, image signal processing IP and network interfaces, all while integrating the machine-learning inference.
The SDK is scheduled to be available in the third quarter of 2020, starting with an early access programme in June. PolarFire FPGAs are in production today.
Headquartered in Chandler, Arizona, Microchip Technology provides smart, connected and secure embedded control products. The company’s products serve more than 120,000 customers across the industrial, automotive, consumer, aerospace, defence, communications and computing markets.