Computer Vision and Pattern Recognition (554SM, 6 ECTS) introduces the foundations of computer vision: how images are formed, how information is extracted from them, and how machines are taught to interpret them. The course covers both sides of the field — the geometric one (image formation, camera calibration, stereopsis) and the recognition one (image processing and multiscale representations, features, learning foundations, convolutional networks and transformers).
Classical methods are treated at full mathematical depth rather than as history: some geometric problems still have efficient non-deep-learning solutions, and many modern architectures echo classical ideas. The emphasis is on understanding over practice.
The course is organized in five parts — 12 theory lectures and 6 lab sessions in Python on Google Colab — and is taught in English. The main reference is A. Torralba, P. Isola and W. T. Freeman, Foundations of Computer Vision (MIT Press, 2024), freely readable at https://visionbook.mit.edu/; A. Fusiello, Computer Vision: Three-Dimensional Reconstruction Techniques (Springer, 2024) is the reference for the geometric part.
Assessment is a closed-book written examination (one third of the grade) and an oral examination (two thirds); there is no project.
Detailed organization of the course:
- Part I — Foundations and image formation (9h)
- L1 the challenge of vision · L2 image formation · Lab1 a simple vision system
- Part II — Image processing and multiscale (12h)
- L3 image processing · L5 sampling and pyramids · L6 Scale-space and SIFT · Lab2 edge detection and multiscale
- Part III — Features and fitting (15h)
- · L7 fitting geometric primitives · L8 camera calibration · L9 stereopsis · Lab3 Estimating homography · Lab4 Estimating fundamental matrix
- Part IV — Learning foundations and CNNs (12h)
- L04 learning foundations · L10 classical recognition (bag-of-words) · L11 neural architectures · Lab5 training a CNN and visualizing the weights
- Part V — Transformers (6h)
- L12 transformers · Lab6 transformers lab
- Teacher: FELICE ANDREA PELLEGRINO