README.md

Vitis™ 2020.2 / Vitis-AI™ 1.3 - Machine Learning Tutorial for the ZCU104 See Vitis™ Development Environment on xilinx.com See Vitis-AI™ Development Environment on xilinx.com

This tutorial is divided in 3 sections.

• Section 1 :
  • An overview of Vitis and Vitis-AI Workflows
    • See how Vitis unifies software, acceleration, and ML development under a single development platform.
• Section 2 :
  • Vitis software platform setup
  • Vitis-AI setup
• Section 3 :

  • Deploy a DenseNet inference application on the ZCU104 board

    • Video file input
    • USB camera input

  • Increase overall system performance by using the Vitis Vision Library to accelerate the image pre-processing

  • Module 1

    • Prepare SD card with the pre-built DPU platform
    • Boot the ZCU104 and verify basic functionality

  • Module 2

    • Setup cross-compilation environment
    • Update glog package
    • Cross-compile the Vitis-AI examples

  • Module 3

    • Update the board image
    • Run RefineNet demo

  • Module 4

    • Classification using Vitis-AI and Tensorflow
    • Running model through the Vitis-AI tool flow
    • Deploying the model to the ZCU104 and evaluating results

  • Module 5

    • Working with network and Vitis-AI
    • Modifying RefineDet model to work with Vitis-AI
    • Train model with modified dataset
    • Use Vitis-AI to generate deployment files
    • Running RefineDet on the ZCU104

  • Module 6

    • Review the Vitis-AI APIs for application development
    • Review the RefineDet application architecture
    • Cross-compiling RefineDet application using the cross-compilation environment

  • Module 7

    • Determining performance bottlenecks in RefineDet application
    • Accelerating the image pre-processing using the Vitis Vision libraries
    • Measuring end-to-end system performance

^{Copyright© 2020 Xilinx}