S-114.204 Havaitsemisen mallintaminen/ Modelling of perception
	HUT / Units / LCE / Teaching / S-114.204

Modelling of Perception

Spring 2005

3 credits, L, 27+27

NEW: Additional exam 26 May - see below.

The course is organized jointly by Helsinki University of Technology and University of Helsinki. University of Helsinki course homepage is here.

The language of the course is English.

Contents

Mechanisms and modelling of human perception are the topics of the course. The emphasis is on modelling of visual feature extraction, and modelling of object and scene recognition. Human visual system and other topics dealing with modelling of visual perception are also included.

Aims

The aim is to give an overview of how visual processes can be modelled and what kind of challenges exist. Particularly, knowledge of modelling of visual feature extraction and object recognition is crucial, not only for the understanding of the human visual system, but also for biologically motivated computer vision applications.

Prerequisites

Basic courses in differential calculus, probability and linear algebra. Also basic knowledge of the human visual system is recommended (e.g. courses Perception and Action, or Structure and Functioning of the Human Brain).

Teachers

Aapo Hyvärinen
Jouko Lampinen
Kaisa Tiippana

Lectures

Thursdays at 14.15-16.00.

Jouko Lampinen's and Kaisa Tiippana's lectures will be given in lecture room H302 at Helsinki University of Technology, Department of Electrical and Communications Engineering, Otakaari 5 A.

Aapo Hyvärinen's lectures will be given in lecture room B222 at University of Helsinki, Kumpula Campus, Exactum building. See here for directions on how to get there.

Literature

Material is in English and consists of lecture notes and journal articles/book sections. The lectures and lecture notes address the topics of the course. Articles and book sections should be read concentrating on those issues. Articles/book sections also give other background information which may be of interest but which is not required. Material is distributed in the beginning of each lecture and can be requested from each lecturer. More details will appear here when the course begins.

Material for Kaisa Tiippana's lectures:

Username and password are required to access handouts and articles, and can be requested from Kaisa Tiippana. (They are the same as for S-114.780.)

• Lecture 1 handout.
• Lecture 1: Palmer (1999). Vision Science. Photons to Phenomenology. Chapter 2. Theoretical Approaches to Vision (pp. 45-93). (Not available in electronic form.)
• Lecture 7 handout.
• Lecture 7: Lee & Mumford (2003). Hierarchical Bayesian inference in the visual cortex. Journal of the Optical Society of America A, 20, 1434-1448.
• Lecture 8 handout.
• Lecture 8: Wallis & Bülthoff (1999). Learning to recognize objects. Trends in Cognitive Sciences, 3, 22-31.

Additional material for the interested (not required):

• Lamme & Roelfsema (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neuroscience, 23, 571-579.
• Hochstein & Ahissar (2002). View from the top: hierarchies and reverse hierarchies in the visual system. Neuron, 36, 791-804.

Material for Aapo Hyvärinen's lectures:

On HU course pages.Username and password are the same as for KT's material.

Material for Jouko Lampinen's lectures:

Use the same username and password as for Kaisa's stuff

For some unknown, probably MS-related issue, the pdf-files of the handouts are huge, 3-9 MB. Sorry for inconvenience.

• Lecture 9 handout.
• Lecture 10 handout.
• Lecture 10: Kersten & Mamassian & Yuille, Object perception as Bayesian inference,Annu. Rev. Psychol. 2004. 55:271-304. .
• Lecture 11 handout.
• Lecture 12 handout.
• See also Wallis & Bülthoff and Lee & Mumford papers in Kaisa's material.

Practicals

• Practical 1 is related to modelling of early vision. It consists of simulating V1 receptive fields using Gabor filters and filters learned from natural images using ICA. The instructor is Jukka Perkiö. Description of Practical 1 is here.
• Practical 2 is related to modelling of object vision. It consists of building a system that can infer the 3-dimensional structure of a wire frame model from 2-dimensional view. The instructor is Jouko Lampinen. Description of Practical 2 is here.

Practicals are compulsory. They are done independently, i.e. there are no specified contact hours, but do not hesitate to ask for advice from the instructors.

Requirements

Written examination

• There are 6 questions to choose from (2 by each lecturer).
• Pick any 4 questions to answer.
• Each question is worth 6 points, so the maximum score in the exam is 24 points.

The first exam is on Tuesday 10 May, 2005, at 9.00-12.00 o'clock in lecture hall S2, Department of Electrical and Communications Engineering, Otakaari 5 A.

The second exam is on Thursday 26 May, 2005, at 14.00-17.00 o'clock in room B317, Laboratory of Computational Engineering, Innopoli 2, Tekniikantie 14.

Register to the exams via wwwTopi,or by e-mail to Kaisa Tiippana (kaisa.tiippana@tkk.fi) one week before the exam at the latest (by 7 May and 19 May, respectively).

Practicals

Practicals have to be passed at an acceptable level and good performance can increase the course mark.

• For each practical you can get 0-3 extra points: 0=just acceptable, 1=good, 2= very good, 3=excellent.
• Each practical thus accounts for 10% of the course mark. Full course score: exam + practical 1 + practical 2 = 24 + 3 + 3 = 30.

Links

• Max-Planck-Institute for Biological Cybernetics Research (Bülthoff's lab)
• The Joy of Visual Perception: A Web Book (basics of vision)
• Neuroscience tutorial
• English-Finnish vocabulary for KT's Perception and Action course

Additional information

Register on the course preferably via wwwTopi,or by e-mail to Kaisa Tiippana (kaisa.tiippana@tkk.fi), or at the first lecture. wwwTopi registration will be closed on 15.2. since there will be a system update.

Recommended for students in their 3rd year or beyond.