AI on industrial standard SOM


Deep Neural Networks, as subsets of Machine Learning and Artificial Intelligence (AI), can efficiently solve complex computer vision problems such as image classification, object detection, image segmentation, and many others.

FPGAs are becoming a major player in the field of embedded AI applications due to their ability to implement complex neural networks with low power consumption and low latency, while simultaneously interfacing large number of peripherals and providing high level of robustness, important for industrial applications.

The Xilinx Vitis AI development environment enables the adaptation and deployment of state-of-the-art neural networks in embedded applications on FPGAs from popular machine learning frameworks such as Tensorflow, Caffe and PyTorch.

To explore the potential of the embedded AI applications, Enclustra adapted one of the embedded real-time image processing applications to run on Enclustra’s own System on Module. The application runs on Mars XU3 module, featuring a Xilinx Zynq UltraScale+ MPSoC device, mounted on the Mars ST3 base board. The AI application supports popular neural networks resnet50 and SSD, for image classification and real-time face detection, respectively. Here, the images are captured with a standard USB camera, connected to the Mars ST3 base board. For higher performance, a MIPI interface can be used, available on the Mars ST3. Moreover, adding actuators such as BLDC or stepper motors is a straightforward task using Enclustra’s Universal Drive Controller IP Core.

We are excited to help showcase the strengths of the FPGAs in the field of AI and help position FPGAs in the front lines of the embedded AI applications.

Employed Technologies

Xilinx® Zynq® UltraScale+®| VHDL | Mentor Graphics ModelSim®
Xilinx DNNDK | C++ | Linux

Involved Enclustra Products

Mars XU3 | Mars ST3