Introduction to Deep LearningLaajuus (5 ECTS)

Objective

This course provides an understanding of the basic mechanisms of deep learning models related with image processing. The students will learn how to develop a VGG-16 model for image recognition and a YOLO v1 algorithm for object detection from scratch. These algorithms use Convolutional Neural Network (CNN), which provides excellent performance in the field of image processing, and the main theme of this course is to learn its mechanism and application methods. This course also provides an understanding of how to use PyTorch, one the most popular framework in deep learning, including how to work with both open datasets and original datasets. In the latter half of the course, the students will work on a simple project in a team and improve their programming skills and the knowledge of deep learning through the development experience.

Content

Day 1: Introduction to PyTorch (popular neural network framework)
- Reviews on Neural Network Basics
- Building and training simple neural networks with PyTorch
- Image recognition using Convolutional Neural Network (CNN) and some tricks

Day 2: Image recognition with deep learning
- Image recognition using VGG-16 network model
- Using pretrained models and transfer learning
- How to use original dataset for image recognition

Day 3: Object detection (1)
- An Overview of object detection problems
- Overview of YOLO v1 algorithm
- Building YOLO v1 from scratch (1)

Day 4: Object detection (2)
- Building YOLO v1 from scratch (2)
- Training YOLO v1 with Pascal VOC dataset

Day 5: Mini-project planning
- Team building and project planning
- Deep learning model and dataset preparation

Day 6, 7: Mini- project
- Carrying out system development with deep learning for each team

Day 8: Presentation session
- Students report their project results in teams

Additional content
How to use the pretrained model will also be introduced in the course.

Qualifications

It is desirable to have completed the contents of Introduction to Machine Learning course. Otherwise, object-oriented programming experience is required, especially Python programming experience is desirable. Basic knowledge of neural networks and machine learning algorithms is also desirable. In this lecture, some basic mathematical expressions such as linear algebra and calculus are used.