Organizer	Dipl.-Inf. Philipp Heise, Dipl.-Inf. Brian Jensen, Dipl.-Inf. Sebastian Klose
Module	IN2106
Type	Masterpraktikum
Semester	SS 2013
ECTS	10.0
SWS	6
Time & Location	Wednesdays 14:00 - 15:30 in 02.09.023 Starting on 17.04.2013
Preliminary Talk / Vorbesprechung	Monday, 28.01.2013 at 11:00 in room 03.07.023 00.12.019
Certificate	successful completion of all exercises
Registration	By email to one of the organizers. Confirmation via TUMonline Registration Closed

Course Materials

Date	Due	Topics	Slides	Assignment Sheet
17.4.2013	8.5.2013	ROS familiarization and Harris corner detection	pdf	pdf
08.05.13	22.05.13	Feature Descriptors and Matching: MOPS, BRIEF and ORB, ROC Challenge	pdf	pdf
22.05.13	05.06.13	Stereo Visual Odometry	pdf	pdf
05.06.13	19.06.13	Dense Stereo	-	pdf
19.06.13		Start of Project Phase	-	pdf

Final Projects

Team	Project	Proposal	Final
A-Team	Stereo SLAM	pdf	pdf
Computer Visionaries	Robust RGBD visual odometry using a particle filter	pdf	-
Global Maximum	Systematic evaluation of binary feature descriptors	pdf	pdf
Grotteneumel	Sequence SLAM for place recognition	pdf	pdf
Wurzel	Robotino - the follower	pdf	pdf

ROC Challenge Results

Team	Bikes	Boat	Graffiti
A Team Total AUC: 0.454 Total PPV: 0.308	AUC: 0.544, PPV: 0.382	AUC: 0.401, PPV: 0.126761	0.424, PPV: 0.298
Computer Visionaries Total AUC: 0.830 Total PPV: 0.197	AUC: 0.933, PPV: 0.436	AUC: 0.863, PPV: 0.098	AUC: 0.752, PPV: 0.075
Global Maximum Total AUC: 0.691 Total PPV: 0.749	AUC: 0.784, PPV: 0.662	AUC: 0.621, PPV: 0.842	AUC: 0.669, PPV: 0.742
Grotteneumel Total AUC: 0.747 Total PPV: 0.503	AUC: 0.644, PPV: 0.397	AUC: 0.774, PPV: 0.708	AUC: 0.823, PPV: 0.534
Wurzel  Total AUC: 0.899 Total PPV: 0.207	AUC: 0.846, PPV: 0.287	AUC: 0.928, PPV: 0.193	AUC: 0.922, PPV: 0.140

Description

This lab course will cover various interesting topics of Computer Vision useful for Robotic applications. The participants will start by implementing some fundamental computer vision algorithms. These basic components will be iteratively expanded as the course progresses resulting in the creation of a full featured visual processing pipeline that in turn can be used to solve complex robotic tasks, in particular SLAM. Participants will then select a specific project from a range of final project topics to further expand their skills and knowledge in desired area of computer vision as the completion of this course.

Selection of possibly covered Topics

• Robot Operating System
• Feature Detectors and Descriptors
• Feature Matching and Tracking
• Visual Odometry
• VSLAM
• 3D Reconstruction
• RGBD-Sensors
• Binocular Stereo

General Information

• The course will be held in English language. The materials will also be in English.
• The number of possible participants is 15.
• The course will be likely be divided into two parts: regular exercise sheets in the first part and a final project in the second part
• Requirements: Working knowledge in C++, basic knowledge in Computer Vision
• See the course materials from the previous semester for an idea about the structure of the course.

External Collaborators

• BMW Car IT GmbH

Literature

• Richard Szeliski - Computer Vision: Algorithms and Applications
• Simon J.D. Prince - Computer Vision: Models, Learning, and Inference