Teaching code

4S009013

Credits

6

Coordinator

Umberto Castellani

Language

English

Scientific Disciplinary Sector (SSD)

INF/01 - INFORMATICS

Courses Single

Authorized

Moodle Seminars 0

The teaching is organized as follows:

Learning objectives

This course is aimed at providing the student with the practical and theoretical tools that enables the recovery of the three-dimensional structure of a scene starting from its two-dimensional projections, the images. Formal methods are provided for image acquisition and extraction of 3D information on several applicative contexts by focusing on the processing of real and noisy data. At the end of the course, the student will be able to implement a new vision system, also in a research context, through the integration of different formal methods on the 3D estimation from images and the use of different acquisition sensors. This knowledge will allow the student to: i) exploit the knowledge of computer vision on different applicative scenarios; ii) master the analysis of real and heterogeneous data; iii) address real time performances. At the end of the course, the student will be able to: i) identify the vision method most suitable to the involved applicative context, and customize the vision system involving other disciplines like machine learning; ii) continue independently his/her studies in the field of computer vision and analysis of 3D data independently.

Prerequisites and basic notions

It is useful to remind the notions of linear algebra, mathematical analysis, and numerical methods. These notions will be reviewed during the course.

Program

------------------------
UL: Teoria
------------------------
- Geometry of the pinhole camera
- Calibration
- Epipolar geometry
- Triangulation
- Planes and homographies
- Structure and motion from images
- Autocalibration
- Dealing with noise and outliers
- Image matching
- Non-rigid objects
- Laboratory exercise
------------------------
UL: Laboratorio
------------------------
- Geometry of the pinhole camera
- Calibration
- Epipolar geometry
- Triangulation
- Planes and homographies
- Structure and motion from images
- Autocalibration
- Dealing with noise and outliers
- Image matching
- Non-rigid objects
- Laboratory exercise

Bibliography

Didactic methods

------------------------
UL: Teoria
------------------------

------------------------
UL: Laboratorio
------------------------

Learning assessment procedures

The exam can be obtained with three different options:
A) Oral with discussion on lab exercise .
B) Project with discussion on lab exercise.
Oral is a discussion on the program. The aim is to verify the knowledge of theoretical and practical aspects of involved topics.
The discussion of lab exercise consists of the delivering of an archive with the scripts that implement the vision algorithms described in the program. The discussion aims at verifying the correct practical implementation of the theoretical aspects addressed during the course.
The project is focused on a specific and innovative topic that is identified with the teacher. The topic can be an open issue of the state of the art or a specific applicative theme. The student will be able to generalize the knowledge acquired during the course for the solution of new computer vision problems.

Evaluation criteria

The discussion of the laboratory exercises represents a barrier (positive or not negative). The oral mark is divided into the following criteria: 18-23: the student knows the basic subject. 24-27: the student links the topics well. 28-30L: the student masters the more advanced aspects. The grade of the project is divided into the following criteria: 18-23: the project is a simple extension of the exercises carried out in class. 24-27: the project involves innovative topics not addressed in the classroom. 28-30L: the project involved state-of-the-art topics.

Criteria for the composition of the final grade

Max 30L

Exam language

Italiano o Inglese