This application note is intended for AI and FPGA developers who want to install and run the FPGA AI Suite Docker* image with a Docker* client running on a Microsoft* Windows* system. The containerized FPGA AI Suite enables easy and quick access to the various tools in FPGA AI Suite.

This document also includes a quick-start tutorial that results in a ResNet-50 network running inference on an SoC FPGA device. The tutorial covers compiling an architecture with FPGA AI Suite compiler, generating an IP file, programming an SoC FPGA device, and running an example inference program on the device.

The quick-start tutorial included in this application note demonstrates the ease of use and ease of deployment of a deep learning model to the edge. The tutorial presents the Memory-to-Memory (M2M) execution model that provides a dla_benchmark interface to the inference engine, similar to the PCIe-based design examples. It also introduces the Streaming-to-memory (S2M) execution model. The S2M model uses the FPGA SoC ARM CPU as the streaming data source that streams data to a layout transform on the FPGA device. The S2M example design illustrates a recommended system architectures for any streaming source.