Deep Learning for Embedded Applications

Machine learning is revolutionizing our lives by automating, simplifying work flows, and improving several industries to our benefit. We see machine learning increasingly becoming an important factor in embedded applications.

This course will enhance your knowledge towards a concrete use of machine learning by providing an introduction of machine learning, and in particular its importance with embedded edge devices. During the course, we will retrain an image recognition model that detects objects. This tutorial includes hands-on exercises, and you will use Google’s Tensorflow software on your own laptop, cloud computing and Raspberry Pi with a camera attached.

PREVIOUS KNOWLEDGE

We use Python on this course. No previous Python experience is required, but understanding basic programming concepts is require. You should also have previous experience using Linux/UNIX as a user.

PRACTICAL EXERCISES / TOOLS

About half of the time is spent on practical exercises. They are designed to emphasize the development process of machine learning process.

We use a PC as a host. On the target we run Linux.

You will keep the latest Raspberry Pi 4 and camera and CORAL AI accelerator so you can continue to study machine learning after training.

CONTENT
INTRODUCTION

• What is machine learning?
• Material and tools used during training.

OVERVIEW

• Basic terminology walk-through
• Process for training and executing
• Industry and ecosystem overview

BASIC MACHINE LEARNING

• Training a model
• Test and validation in machine learning
• Basic tuning of a model

DEEP LEARNING

• What is deep learning
• Convolutional neural networks
• Recurrent neural networks
• Transfer learning

INFERENCE PERFORMANCE ON EDGE DEVICES

• Reducing accuracy of the model in exchange of performance
• Performance comparison of different computing hardware

FURTHER LEARNING WHAT TO DO AFTER THIS COURSE

• Intermediate terminology and concept walk through to aid in further learning in the field