Description

About this course

Computer Vision is the art of distilling actionable information from images.

In this hands-on course, we’ll learn about Image Analysis techniques using OpenCV and the Microsoft Cognitive Toolkit to segment images into meaningful parts. We’ll explore the evolution of Image Analysis, from classical to Deep-Learning techniques.

We’ll use Transfer Learning and Microsoft ResNet to train a model to perform Semantic Segmentation.

What you’ll learn

• Apply classical Image Analysis techniques, such as Edge Detection, Watershed and Distance Transformation as well as K-means Clustering to segment a basic dataset.
• Implement classical Image Analysis algorithms using the OpenCV library.
• Compare classical and Deep-Learning object classification techniques.
• Apply Microsoft ResNet, a deep Convolutional Neural Network (CNN) to object classification using the Microsoft Cognitive Toolkit.
• Apply Transfer Learning to augment ResNet18 for a Fully Convolutional Network (FCN) for Semantic Segmentation.

Prerequisites

• Working knowledge of Python
• Skills equivalent to the following courses
• Introduction to AI
• Deep Learning Explained

Estimate Time : 12-16 hours

Module 1 Introduction

• The Evolution of Computer Vision
• Image Processing Basics
• Lab

Module 2 Image Features and Classical Method

• Thresholding
• Clustering
• Region Growing
• Template Matching
• Edges and Corners
• Lab

Module 3 Object Classification and Detection

• Viola-Jonas
• HOG
• Classical VS Deep
• Deep Learning
• Classifiers to Detectors
• Object Proposal
• CNN Object Detectors
• Lab

Module 4 Deep Segmentation and Transfer Learning

• Super-Pixels and Conditional Random Fields
• Fully Convolutional Approaches
• Deep Segmenters
• Transfer Learning
• Lab

Andrew Byrne
Senior Content Developer Microsoft Corporation

Andrew is a Senior Content Developer at Microsoft. His passion for software and teaching comes from 20+ years of software development experience at Microsoft, Siemens, Ericsson and his own startup.

Ivan Griffin, PhD
Founder Emdalo Technologies, Ltd.

Ivan Griffin is a director and founder of Emdalo Technologies, where he works on developing embedded machine learning solutions. Ivan has over 20 years of experience in the embedded and semiconductor industries. He has a strong technical background combined with commercial and strategic understanding, and a proven track record in a number of successful start-ups. He has co-authored one patent application in computer vision, and two European and US patents in digital broadcast radio. Ivan has a Bachelor’s (1995) and Master’s degree in Electronic/Computer Engineering (1997) and Ph.D. (2010) in Computer Science from the University of Limerick, Ireland.

Daire McNamara
Founder Emdalo Technologies, Ltd.

An engineer by training, Daire co-founded Emdalo Technologies in 2013 with Dr. Ivan Griffin to realize Machine Learning at the Edge. Daire has over 20 years’ experience in the high-tech electronics industries, having held senior commercial, management and product development roles in start-up and early phase companies targeting US, Asia-Pacific and European markets.

Additional information

Certificate