Training course: Machine learning for embedded vision

Course Objectives

“Machine learning for embedded vision”, is three-day hands-on course, will give you a complete overview of the different tools, frameworks and concepts of machine learning. In just three days, you will clearly understand the basic techniques of machine learning, mathematical concepts and engineering solutions for everyday use.

  • Understand the constraints of programming machine learning algorithms
  • How to manage and process deep learning on embedded target
  • What is the best target? CPU, GPU, FPGA,…

Who Should Attend?

  • Engineers who want to get familiar with basic principles of machine learning applied to computer vision
  • People who want to become fully skilled ML engineers without wasting months on it

Course outline

  • Understand deep learning concepts: feature extraction, neural networks, back-propagation… Lab #1: Training my first classifier
  • Applications and use-cases for computer vision Classification, detection and segmentation Lab #2: Building a real computer vision app.
  • Recipes of efficient deep learning Methods to accelerate deep learning execution: pruning, quantization and complexity reduction
    Lab #3: Deploying my deep learning algorithm
  • Hardware for embedded deep learning Hardware architectures to accelerate deep learning execution: GPUs, ASICs and FPGAs Lab #4: Enhancing the real-time performance of my embedded computer application

Prerequisite

This course request the following prior knowledge:

  • Knowledge of the basis of image processing (filters, convolution, …)
  • Good knowledge of HDL languages (Verilog, VHDL, …)
  • Good knowledge of C/C++ programming

Language

This training course can be taught in English, French, Italian and Arabic