Università degli Studi di Napoli "Parthenope"

Teaching schedule

Academic year: 
2018/2019
Belonging course: 
Course of Master's Degree Programme on APPLIED COMPUTER SCIENCE (MACHINE LEARNING AND BIG DATA)
Location: 
Napoli
Disciplinary sector: 
INFORMATICS (INF/01)
Language: 
Italian
Credits: 
6
Year of study: 
2
Teachers: 
NARDUCCI Fabio
Cycle: 
First Semester
Hours of front activity: 
48

Language

Lectures in Italian (and in English on e-learning), books in English.

Course description

The aim is to provide the notions of the camera pinhole model and the geometric relationships that exist between multiple images of the same
scene, recent video analysis techniques, such as Bag of Words and those based on weaving and symmetry measurements, but above all the role of
Machine /Deep Learning in Computer Vision.

Knowledge and understanding:
Student should demonstrate knowledge and understanding of the basic theoretical and practical aspects for video analysis problems (video analytics) such as video classification, video summarization, video indexing, with particular attention to extraction and learning of video information.

Ability to apply knowledge and understanding:
Student should demonstrate that he/she can use his/her own acquired knowledge to analyze complex and large video datasets with the aim of knowing how to apply and parameterize the most recent techniques and algorithms for video analysis.

Making judgments:
Student should be able to evaluate in autonomously way the complexities inherent in video datasets and be able to identify and evaluate the
strengths and weaknesses of the known machine / deep learning algorithms to solve classification problems and clustering on video.

Communication skills:
Student should be able to argue with complex scientific rigor and appropriate terminology the complex concepts of machine / deep
learning techniques as well as knowing how to work out one critical evaluation of the results deriving from experimentation on cases real studies.

Learning skills:
Student should be update himself/herself on the most emerging machine learning and deep learning approaches in computer vision community. It
must also be able to deepen into autonomous way the themes presented during the course in order to elaborate a personal vision of the state of the art.

Prerequisites

A good knowledge of mathematical analysis is required to understand techniques and algorithm discussed during the course.

Syllabus

The extended program of the course is organised in the following lessons:
1. Synergies and divergences between image processing and
computational vision
2. Geometry of image formation
3. Camera pinhole model and calibration
4. Epipolar Geometry and Rectification
5. Relative pose, Essential matrix: calculation and factoring
6. Absolute pose
7. External pose
8. "Point-based" reconstruction
9. "Camera-based" reconstruction
10. Texture analysis
11. Texture synthesis
12. Symmetries in 3D images
13. Recurrence/Feedback based image representations
14. Structured image representations
15. 2D and 3D Object and scene representations
16. Unsupervised learning of representations

The course aims to provide the student with skills and knowledge related to Computer Vision. Computer Vision consists in deducting the properties of the world based on one or more digital images through algorithms for the analysis of images and videos based on color, texture, shading, stereo and movement. It includes the study of the camera pinhole model and of the geometric relationships that exist between multiple images of the same scene, recent video analysis techniques, such as Bag of Words and those based on weaving and symmetry measurements, but above all the role of Machine / Deep Learning in Computer Vision.

Teaching Methods

The teaching consists of a total of 48 hours (6 CFU). Teaching consists of theoretical and practical lessons on the topics covered by the course. It takes place in the classroom or in the laboratory and is organized in 2 hour lessons marked by the academic calendar.

The laboratory lessons are aimed to address to already reported codes on the web related to the theoretical topics seen in class discussing about the effective algorithmic implementation and highlighting its potential and the limitations on different video datasets available online. Course attendance is optional. The exam tests will be the same for all students, attending or not.

Textbooks

B. Forsyth & Ponce, Computer Vision: A Modern Approach (2nd Edition), Prentice Hall, 2012, ISBN-10: 013608592X, ISBN-13: 978-0136085928
The book is accessible at link
http://cmuems.com/excap/readings/forsyth-ponce-computer-vision-amodern-a...

Lecture on e-learning http://e-scienzeetecnologie.uniparthenope.it over
the topic Deep Learning given from Prof. Petrosino

Learning assessment

The exam consists in producing a report for evauation. The examination consists in the evaluation of the paper report on the basis of clarity and
completeness.
Students will be asked to study an article of their choice among research papers in the Open Access versions, provided by the Computer Vision Foundation and available at link http://openaccess.thecvf.com/menu.py and produce a written report to be submitted / sent to the teacher for
evaluation.
The report can be only related to therorethical aspects or to a code available on web, consistent in a sequence of analysis steps of the images that lead to the interpretation of the image, thus underlining the experience.

It is possible to work in a group of 2 students and produce a single report. The student can also deliver / send in an optional way and at his own
discretion a Power Point presentation (or similar) with recorded audio related to the paper report.
Possible structure of the report:
Introduction:
introduce the topic, provide a brief overview of the topic.
Report with previous topics:
Briefly describe the technical aspects of the main results in the area.
Review of previous work:
This is an important section. Explain the details of the method and the
technical solution proposed by the authors of the selected scientific work.
Technical part:
Summary of the adopted technical solutions in the selected scientific
work, followed by details of technical solutions. Discuss the similarities /
differences of scientific works within the area when appropriate. If
relevant, try to show how the individual works are based on other jobs in
the area.
Results:
Present the experimental results of different methods with graphics,
images and visualizations taken from the literature or related to a specific
software already available on the web.
Conclusions:
What is the message to take at home?
References:
Cite the most relevant works, many already present in the selected work,
to which add possibly those considered useful for a better understanding
of the produced report.

More information

The learning material is provided, in English, through the e-learning platform at the following weblink
http://escienzeetecnologie.uniparthenope.it. To access the published material, the student must be regularly registered at the platform.