Organizer |
Chavdar Papazov, M.Sc. |
Modul |
IN8901 |
Type |
Hauptseminar |
Language |
English |
Term |
Winter 2009/2010 |
ECTS |
4.0 |
SWS |
2 |
Preliminary Meeting |
Thursday, July 23, 2009 14:00 MI 03.07.023 |
Time & Room |
Tuesday 14:00 - 15:30 MI 03.07.023 |
News
Mon Oct 12, 2009: The presentation dates are fixed now (see the table below).
Tue Oct 7, 2009: All papers are assigned (see table below). There are no more free places for the seminar.
Wed Sep 30, 2009: The time slot for the seminar is now fixed (check the table above). There will be 3 meetings each consisting of two presentations. The exact dates will be announced soon.
Thu Sep 17, 2009: The title of the Hauptseminar has been changed from "Model-Based 3D Object Recognition" to "3D Object Recognition and Registration".
Overview
Object Recognition is important in many practical applications of computer vision. The traditional techniques based on 2D images are sensitive to changes in illumination, shadowing and viewpoint. When a 3D description of objects and scene is available, the recognition procedure does not have to deal with these problems.
The aim of 3D object recognition is to correctly identify known objects in a scene and estimate their pose. 3D models of the objects are constructed offline and are stored in a model library using a suitable representation. During online recognition, the input scene is converted into a similar representation and matched with the models from the library.
Object Registration is the automatic alignment of partially overlapping datasets (objects) such that they build a complete, consistent model. It can be considered as part of the object recognition, since registration is involved in the pose estimation step of the recognition process.
The topics to be presented and discussed in this seminar are related to:
- Interest point detection: detection of salient points (features) from the input sets, for example, corner points.
- Local geometry description: representations which characterize the geometry of the input data in the vicinity of feature points.
- Registration: determining of a set of transformations which align the data sets (in some optimal way) to each other.
How to participate?
Please send an email to Chavdar Papazov (
papazov@in.tum.de) if you are interested in the seminar and would like to participate. Provide the following information
- Surname, Name
- Matriculation number
- Branch of study
The number of participants is limited to 6.
Presentations/Material
Each student will give a presentation based on a single paper. The presentation should be 20 to 30 minutes followed by 5 to 10 minutes for questions and discussion. Send your slides to Chavdar Papazov (
papazov@in.tum.de) a week before your talk takes place.
The slides from the preliminary meeting can be downloaded
here.
Date |
Title |
Speaker |
Paper |
Advisor |
Nov 17, 2009 |
A Method for Registration of 3-D Shapes (ICP) |
Stefan Kinauer |
PDF |
Chavdar Papazov |
Nov 17, 2009 |
Geometric Hashing: A General And Efficient Model-Based Recognition Scheme |
Zlatina Cheva |
PDF |
Chavdar Papazov |
Nov 24, 2009 |
Using Spin Images for Efficient Object Recognition in Cluttered 3D Scenes |
Yanko Sabev |
PDF |
Chavdar Papazov |
Nov 24, 2009 |
4-Points Congruent Sets for Robust Pairwise Surface Registration |
André Aichert |
PDF |
Chavdar Papazov |
Dec 15, 2009 |
Robust Global Registration |
Anastas Tanushev |
PDF |
Chavdar Papazov |
Dec 15, 2009 |
Salient Geometric Features for Partial Shape Matching and Similarity |
Dimiter Dimitrov |
PDF |
Chavdar Papazov |
Grading
In order to get the credits (ECTS/Schein), you have to give a presentation, write a summary and attend all seminar meetings. The summary of your talk should be 5 to 10 pages. Hand it in by the end of the year 2009 (but better finish your summary before you give your talk, because trying to write things down in your own words will help you realize which parts of the paper(s) are important).