Deep Learning Based Machine Vision

Fast development and wide application of algorithms and systems for machine vision in the last decade is mostly based on achievements in the domain of neural networks and deep learning. The goal of the course is study of the neural networks architectures that are used in machine vision systems, with the focus on image classification, object detection in images and image segmentation. The students will study the theory of neural networks, their architecture and implementation. Upon successful completion of the course the students will be able to design neural networks architectures for solving specific problems in the domain of machine vision, to implement them, to train and test them on real world problems. They will acquire knowledge about new developments in the field and will be able to further broaden their knowledge and skills through implementation of their own research.

Introduction to neural networks, perceptron, activation function, loss function, multilayer neural networks, training and backpropagation, practical issues in neural networks training, deep neural networks, recurrent neural networks, convolutional neural networks (CNN), CNNs for image classification, CNN based object detection, CNNs for segmentation, attention and transformers, video analysis and 3D CNNs.

Combined learning methods: lectures, supported by presentations and concepts visualizations, active participation of students through project works.