COMPUTER VISION
Faculty of Engineering - University of Pavia - a.y. 2017-2018

Prof. Virginio Cantoni



Please visit the facebook and web page of the Computer Vision and Multimedia laboratory (CVMLab).


Last update, 25 June 2018

This course is intended for advanced undergraduate students. We assume students have a rudimentary understanding of linear algebra, calculus, and are able to program in some type of structured language.

Computer Vision consists of inferring properties of the world based on one or more digital images. Provides background in image processing and image formation. Focus on algorithms for image and video analysis based on color, texture, shading, stereo, and motion.

Contents

Introduction to Computer Vision
Basic definitions. Low-level image analysis methods, including image formation, edge detection, feature detection, and image segmentation.
3D Vision and motion analysis
Methods for reconstructing three-dimensional scene information using techniques such as depth from stereo, structure from motion, and shape from shading. Motion and video analysis.
Object recognition
Recognition Processes. Direct Comparison. Alignment methods. Invariant properties methods. Parts decompositions method. Hough transform.
Image synthesis
Computer graphics topics involving computational photography and image-based rendering. Local rendering, Phong model. Advanced rendering techniques, topics include ray casting, ray tracing, and radiosity.



Lesson 1 - Introduction

Lesson 2 - The photometric track

Lesson 3 - Spatial track

Lesson 4 - Features and detectors, by Prof. Emanuel Aldea (University Paris XI, Paris-Sud - France)

Lesson 5 - Robust estimation, by Prof. Emanuel Aldea (University Paris XI, Paris-Sud - France)

Lesson 6 - Two-view geometry, by Prof. Emanuel Aldea (University Paris XI, Paris-Sud - France)

Lesson 7 - Multi-view geometry, by Prof. Emanuel Aldea (University Paris XI, Paris-Sud - France)

Lesson 8 - Morphological track

Lesson 9 - Shape extraction

Lesson 10 - Visual search and Hough Transform

Lesson 11 - Spatial track: motion modeling

Lesson 12 - Mathematical Morphology 2D


Codici esempio:
- C-utilities
- Java-utilities (updated 22/10/2015)
- example images in pgm format (updated 22/10/2015)
- example colour images


KIRO page - The course on the eLearning platform of the University of Pavia

Bibliography

Book: "3C Vision - Cues, Contexts and Channels", V. Cantoni, S. Levialdi, B. Zavidovique, Elsevier, 2011
course slides (please download at this website address)

Information on the Project Structure

- choose one programming language: C, C++, or Java
- do not use toolbox and packages for image processing
- work in a group of 2 people
- write a report (Word or HTML)
- the project can be submitted on cd-rom; please write name and surname of each participant and the academic year
- the project will be evaluated according to: clearness, completeness, results and related discussion
- Structure:
  1. objective of the project
  2. description of the implemented theory including images, drawings, etc.
  3. insert images before and after the elaboration
  4. comment the results regarding efficiency, precision, computing time, comparison with other solutions, etc.
  5. insert the source code ANNOTATED and compiled with all the necessary to run as Appendix


TEST modalities

Academc year 2017/2018
Students will be asked to read three papers. They will be required to write a report of one of these papers, due before we discuss the paper. Each student is required to complete a laboratory project consisting of a sequence of image analysis steps resulting in image interpretation thus emphasizing hands-on image analysis experience. The exam consists on the discussion of the projects and on the paper report.

Each student will give a Power Point (or similar) presentation on a paper taken from the International Conference on Computer Vision (ICCV 2017) October 22-29, 2017, Venice, Italy.
Each group component must select a different paper among five possible (there is no priority even if I called them first, second, .. , fifth choice: I selected the short list just based on our course content and the alphabetic order of the accepted papers at the conference).
The assignments are given by the sequence of group names number (see here) and the paper sequence in the following five folders:
http://vision.unipv.it/CV/materiale2017-18/firstchoice,
http://vision.unipv.it/CV/materiale2017-18/secondchoice,
http://vision.unipv.it/CV/materiale2017-18/thirdchoice,
http://vision.unipv.it/CV/materiale2017-18/fourthchoice,
http://vision.unipv.it/CV/materiale2017-18/fifthchoice.


Characters to be modeled in 3D - assignments to each student

Here is a project description and a related paper.

Please find material for the Final Homework 2018:

L'arcangelo Raffaele che guida Tobia, con il pesce ed il cane
L'Arca di Sant'Agostino, Pavia e dintorni

Additional pictures: from L'Arca di Agostino, Agostino e la sua Arca, other_1, and other_2 - uploaded Dec 07, 2017
More pictures_1 and More pictures_2
Arca
"la Città dell'Uomo"
"Agostino e la sua Arca".


Software tools:
Blender is a professional, free and open-source 3D computer graphics software toolset used for creating animated films, visual effects, art, 3D printed models, interactive 3D applications and video games. download
Adobe Fuse CC is a 3D computer graphics software developed by Mixamo that enables users to create 3D characters. Its main novelty is the ability to import and integrate user generated content into the character creator. For a limited time, Fuse will be available as a free beta version for members of Creative Cloud. download
CINEMA 4D is a 3D modeling, animation, motion graphic and rendering application developed by MAXON Computer GmbH in Germany. Please ask for CINEMA 4D Studio R17 + upgrade to Studio R18 to Alessandra Setti (alessandra.setti@unipv.it - Computer Vision and Multimedia Lab - CVML).

Adobe Fuse + Blender WORKFLOW, by Federico Guerra and Marco Provitina (e-mail addresses: fede23_caos@live.it, marco.provitina@gmail.com)
Material.blend is the file in which you will find the three materials that you can need for your work. If you open that file, you'll find the instruction written on the blender's text editor.
Blender shortcuts is a pdf file that contains all the most useful shortcuts.


Academic year 2016/2017
At the exam each student will give a Power Point (or similar) presentation on a paper taken from the 23rd International Conference on Pattern Recognition (ICPR 2016) December 4-8, 2016, Cancun, Mexico.
Each group component must select a different paper among five possible (there is no priority even if I called them first, second, .. , fifth choice: I selected the short list just based on our course content and the order submission of accepted papers at the conference and attached workshops).
The assignments are given by the sequence of group names number (see http://vision.unipv.it/CV/Studenti%20elenco.pdf) and the paper sequence in the following five folders:
http://vision.unipv.it/CV/materiale2016-17/firstchoice,
http://vision.unipv.it/CV/materiale2016-17/secondchoice,
http://vision.unipv.it/CV/materiale2016-17/thirdchoice,
http://vision.unipv.it/CV/materiale2016-17/fourthchoice,
http://vision.unipv.it/CV/materiale2016-16/fifthchoice.

For example:
group n. 6, Stefano Di Carlo and Michele Semperboni may select each one one paper (they must be different) from the following list:
0190, 197, 213, 216, 171

Groups - final assignments 2017

Cartella Immagini
Cartella Immagini Final Homework 2017
ELENCO studenti, con assegnazione compito


Academic year 2015/2016
At the exam each student will give a Power Point (or similar) presentation on a paper taken from the 21st International Conference on Pattern Recognition (ICPR 2012) November 11-15, 2012, Tsukuba, Japan, or from the International Conference on Image Analysis and Processing (ICIAP 2015) September 7-11, 2015, Genoa, Italy.
Each group component must select a different paper among six possible (there is no priority even if I called them first, second, .. , sixth choice: I selected the short list just based on our course content and the order submission of accepted papers at the two conferences).
The assignments are given by the sequence of group names number (see http://vision.unipv.it/CV/materiale2015-16/StudentiCV-15-16 progetti.xlsx) and the paper sequence in the six folders:
http://vision.unipv.it/CV/materiale2015-16/firstchoice,
http://vision.unipv.it/CV/materiale2015-16/secondchoice,
http://vision.unipv.it/CV/materiale2015-16/thirdchoice,
http://vision.unipv.it/CV/materiale2015-16/fourthchoice,
http://vision.unipv.it/CV/materiale2015-16/fifthchoice,
http://vision.unipv.it/CV/materiale2015-16/sixthchoice.

For example:
group n. 12, Burda and Firas may select each one one paper (they must be different) from the following list:
"Active Transfer Learning for Multi-view Head-pose Classification",
"A Splitting Algorithm for Directional Regularization and Sparsification",
"Object Segmentation in Multiple Views without Camera Calibration",
"Hierarchical Image Segmentation Relying on a Likelihood Ratio Test",
"Difference-Based Local Gradient Patterns for Image Representation",
"A Survey on Traffic Light Detection",
respectively as first, second, third, fourth, fifth and sixth choice.


Contact: virginio.cantoni@unipv.it (prof. Virginio Cantoni)