Dominik Farhan

Machine Learning in Computer Vision


The notes

These notes were written together with Júlia Križanová. I cannot guarantee they are correct. However, I think they are pretty exhaustive and studying them should be enough to get an excelent grade in the exam.

In multiple places, we added our own bits of ML knowledge. These things might not be important. The exam is write-all-you-know. I thus think that it is not necessary to know perfectly everything covered in the lectures.

All the slides and snippets are from presentations by Elena Šikudová. Any mistakes are solely mine. If you find anything that shouldn't be there, let me know at

Promised notes PDF.

D part of the exam

This part is not covered in the notes. In my experience it is sufficient to study the elementary examples presented during lectures.

My review of the course

Both lectures were quite friendly. You definitely don't need to know any MATLAB. All excersises can be done in Python. Some skill with numpy and previous exposure to scikit-learn make the course much easier although you can do without it.

Overall, the content of the course is very standard and for some reason there are multiple courses at the faculty that cover same things. And I mean literally same things. Also this course is not about computer vision altough it's in the name.

The practicals were finished by two big home works that might take a few days to complete. I liked them both. Not often you'll get homeworks that force you to implement something non-trivial. However, be prepared that it's not one hour thing and you might have to grind it out at the end of your examination period.