Abstract

Artificial Neural Networks and Machine Learning in the form of Deep Learning have proven to be able to solve complex computer vision problems efficiently. This includes image classification, object detection, segmentation and many other applications.

The Xilinx Deep Neural Network Development Kit (DNNDK) for embedded artificial intelligence applications enables the adaptation of state-of-the-art neural networks to FPGAs efficiently and supports popular frameworks such as Tensorflow and Caffe.

By utilizing a standard System on Module (SOM) like the Mars XU3 from Enclustra, that is based on the Xilinx Zynq UltraScale+ MPSoC, in combination with the readily available Mars ST3 baseboard, industrial applications can be implemented in as short a time as never before.

To showcase the power of this combination and the fast time-to-market, Enclustra has developed an AI based image recognition system only in days. The images are captured with a standard USB camera, that is connected to the Mars ST3 baseboard. For higher performance a MIPI interface is also available on the baseboard.

The neural network – that is classifying the images with a low latency – is running on the Mars XU3 SOM. The system is supporting popular neural networks such as resnet50 and SSD for image classification or real-time face detection respectively.

The single FPGA module is not only running the neural network inference, but can also handle numerous other tasks in parallel, like communication with a host PC as well as other peripherals. Moreover, controlling all kinds of high dynamic actuators at the same time is where the FPGA technology is playing to its strengths. For example, adding the Enclustra Universal Drive Controller IP Core to control BLDC or stepper motors would be a snap.

Employed Technologies

Involved Enclustra Products