Deep Learning with visual recognition

The best way to learn about deep learning

Visual recognition (or computer vision) is the most established field within deep learning and also the best way to learn about it. During this 2-day course, you will learn to implement, train and debug your own neural networks and gain a detailed understanding of neural network architectures and insight into cutting-edge research in deep learning.

The hands-on programming exercises will involve setting up computer vision problems, like:

• Image classification
• Apply learning algorithms
• Practical engineering tricks for training and fine-tuning the networks.

At the end of the course we will discuss how to apply deep learning for other data types, like text, speech, and tabular data.

The course is for software developers or engineers, who wish to add deep learning to their toolbox

Participants will be reading and writing code in Python. Basic programming skills are required, but not necessarily familiarity with Python. No prior knowledge about machine learning is required and only basic skills in mathematics are required.

At the end of the course, you will be able to:

• identify and describe visual recognition tasks that can be solved with deep learning
• describe and compare different neural networks architectures,
• explain and relate techniques for training neural networks,
• apply deep learning to standard visual recognition tasks and assess the results,
• define and scope your own deep learning projects.

Your organization will:

• have a competitive advantage by having employees with deep learning competences
• be better equipped to collect and exploit data in the future
• have a higher chance of retaining employees with an interest in AI.

How the course is run

The course covers deep learning theory but focuses primarily on being able to use the techniques and offers hands-on training under expert guidance. You can ask about everything from how to start to when you are stuck. And the best part is that after taking the course you will be able to continue using deep learning at work.

Day 1

• Machine learning fundamentals
• Logistic and linear regression
• Simple neural networks
• Training neural networks (including topics like optimization, backpropagation, transfer learning, and regularization techniques to avoid overfitting)

Day 2

• Convolutional neural networks
• Advanced neural network architectures (including topics like ResNet, Recurrent Neural Networks and Generative Adversarial Networks)

Before the course:

You should either have or be willing to create a Google (i.e., gmail) account, preferably before attending the course.

Also, you will get the most out of the course if you brush up in advance on:

• basic linear algebra (e.g., inner products and matrix/vector multiplication)
• basic calculus (e.g., differentiation and especially the chain rule)
• simple probability theory (e.g., probability distributions)

During the course

You are expected to bring your own laptop to the course.

There are no hardware requirements (like GPU), and you don’t need to install any special software. All you need is an internet browser (ideally Google Chrome).

After the course

There are no hand-ins or anything like that, but you can consult the teacher after the course in case you need help setting up your own projects.