Teaching code

4S008907

Credits

6

Coordinator

Gloria Menegaz

Language

Italian

Scientific Disciplinary Sector (SSD)

INF/01 - INFORMATICS

Moodle Seminars 0

The teaching is organized as follows:

Learning objectives

The course aims at providing competence about analysis, modeling and interpretation of multidimensional signals and images with focus on artificial vision and machine learning aspects, targeting applications in the field of multimedia and interpretable machine learning. At the end of the course the students will be able to autonomously solve typical problems requiring multidimensional signal modeling, feature extraction, analysis and interpretation of the outcomes of machine learning algorithms in the field of multimedia and artificial vision.

Prerequisites and basic notions

Foundations of signal and image processing.

Program

The course aims at providing competence about analysis, modeling and interpretation of multidimensional signals and images with focus on artificial vision and machine learning aspects, targeting applications in the field of multimedia and interpretable machine learning. At the end of the course the students will be able to autonomously solve typical problems requiring multidimensional signal modeling, feature extraction, analysis and interpretation of the outcomes of machine learning algorithms in the field of multimedia and artificial vision with particular focus on convolutional neural networks.

Syllabus
The course consists of three blocks: modeling of the Human Visual System (HVS), multiresolution signal representation and analysis of deep learning algorithms with focus Convolutional Neural Networks (CNNs).

Part 1: Human Visual System (HVS) – 10 hours
Introduction to Visual Intelligence
Foudations of vision, stimulus encoding, representation and interpretation
HVS modeling: multiscale processing of the visual stimuli, Contrast Sensitivity Function (CSF), color vision and perception, Color Matching Functions (CMFs)
High-level modeling of the HVS: structural and functional connectivity and graph-based modeling

Part 2: Multiresolution analysis – 20 hours
Background
Mathematical tools
Fourier transform in 1D and 2D
Windowed Fourier Transform
Wavelets and multiresolution representations
Wavelets Bases
Families of Wavelet Transforms (WT) and their properties
Fast Discrete Wavelet Tranforms (DWT)
WT in two dimensions
Scattering transform

Part 3: Application to the analysis and interpretation of deep convolutional neural networks (CNNs)– 10 hours
Overview on CNNs
The issue of interpretability, main approaches
CNN, HVS and multiresolution: getting to a unified view
CNN interpretation based on multiresolution theory and HVS models
Examples of interpretable DL

LABORATORY
Laboratory sessions will consist in Matlab and Python exercises on the topics covered in the theory lessons.

Bibliography

Didactic methods

The lessons will be both face to face and in streaming and will be recorded and made available on moodle.

Learning assessment procedures

The exam will consist of a project and a colloquium on the topics of the course.

Evaluation criteria

To pass the exam, the student must prove:
- The comprehension of the theoretical aspects at the ground of the three parts of the course;
- To master the links across the topics and the common and different aspects implied in modelling:
- To have a solid background in the multiresolution theory and the understanding of the implications in terms of interpretability of deep learning models;
- The ability to exploit the acquired knowledge for solving concrete multidisciplinary problems.

Criteria for the composition of the final grade

The final score is the average of the scores of the theoretical and practical parts.

Exam language

Italiano