Accelerated Image Classification via Binary Neural Networks Demo

This design example demonstrates how software-implemented neural networks and programmable logic can be accelerated using binary neural network (BNN)s. The demo shows Zynq® UltraScale+™ MPSoC based boards accelerate image classification over 6,000 (Ultra96) to 8,000 (ZCU102) times more than CPU-only. Users can view metrics, images, and classification results via the graphical user interface. Real-time classification may be done via a stored database of images or from a USB camera.