Artificial Intelligence for Computer Vision

Recommended Prerequisites

Linear algebra and analytic geometry, Matrix Calculus, Differential Calculus, Stochastic Estimation, Computer Vision, Signal Processing.

Teaching Methods

The teaching methodologies in this course are based on the presentation of fundamental and conceptual aspects of computer vision in magister classes (theoretical classes), including presentations by students (based on book chapters and papers), complemented with a strong laboratory component where students are able to implement and validate the concepts presented and learened learned on the master classes. The algorithms and applications are implementations based on Matlab.

Learning Outcomes

This curricular unit covers advanced topics in computer vision, covering the study of computer vision from a classical perspective to solutions based on deep learning. The objective of the curricular unit is to provide students with the knowledge and skills necessary to carry out research & development in computer vision and its application domains, such as robotics, biometrics, video surveillance, biomedical applications and graphics. Students must understand the strengths and weaknesses of current approaches to researching problems and identify challenges and future directions for research and development, combining classic computer vision approaches with recent deep learning approaches.
It also aims to improve students' critical reading and communication skills in the fields of computer vision and its applications.

Work Placement(s)

No

Syllabus

1. Detection of characteristic elements and correspondences: Classical approaches (SIFT, SURF, HOG, MSER, FAST, BRIEF, etc.) and methods based on Bayesian approximations and deep learning.
2. object detection, recognition and classification: classical approaches and approaches based on deep learning;
3. Facial Analysis: detection, recognition and verification. Facial action recognition. Classical and deep learning-based approaches;
4. Visual detection and tracking in the image. Tracking multiple objects. Classical and Deep Learning-based approaches.
5. Recognition of activities in video footage.

Head Lecturer(s)

Jorge Manuel Moreira de Campos Pereira Batista

Assessment Methods

Assessment
Other: 10.0%
Research work: 10.0%
Exam: 30.0%
Laboratory work or Field work: 50.0%

Bibliography

"Computer Vision: Algorithms and Applications", 2^nd  editionm Richard Szeliski, Text in Computer Science, Springer, 2022.

"Visual Object Recognition", K. Grauman, B. Leibe, Synthesis Lectures on Artificial Intelligence and Machine Leaerning, Springerr Cham, 2021.

"Dive into Deep Learning", A. Zhang and Z. C. Lipton and Mu Li and A. Smola, arXiv preprint arXiv:2106.11342,2021. Opeen Source access: https://pt.d2i.ai

“Deep Learning”, I. Goodfellow, Y. Bengio, A. Courville, MIT Press, 2016.

"Face Detection and Recognition: Theory and Practice (1^st ed.), A. K. Datta, M. datta, P. Banerje, Chapman and Hall/CRC, 2015.

"Deep Learning-based Face Analysis", N. Ratha, V. Patel, R. Chellapa, Advances in Computer Vision and Pattern Recognition Series, Springer Cham, 2021

"Video Tracking: Theory and Practice," E. Maggio, A. Cavallaro, Wiley, 2011.

"Deep Learning for Computer Vision - Image Classification, Object Detection and Face Recognition in Python, J. Brownlee, Machine Learning Mastery, 2019.