The course offers an introduction to learning approaches used in the field of content-based image retrieval (CBIR). The course is divided into two main parts.
In the first part we consider various forms of image representation. It goes into the difficulties of obtaining automatic image descriptions, and explains where machine learning methods can help in this process. A number of learning methods are treated in detail and we also aim to provide some useful insight in the mathematical theory of classification.
In the second part of the course we discuss learning methods used in the interactive processes in image search systems. We explain why learning during the interaction is crucial, and describe a number of concrete methods to implement relevance feedback on the results presented by image search engines. Finally, we discuss various strategies to evaluate the performance of learning methods.