Exjobbsförslag från företag

Detta är ett uppsatsförslag hämtat från Nationella Exjobb-poolen. Klicka här för att komma tillbaka till samtliga exjobbsförslag.

Förslaget inkom 2010-03-16

Do I know this place? - Novelty Detection for Visual Place Classification

Much of the current research in computer vision, artificial intelligence and robotics concentrates on designing and building cognitive robots able to interact with humans and man-made environments. The ability to represent the environment in which the robot operates in a way which will not only allow for reliable navigation, but will provide higher-level, semantic space interpretation becomes key in such scenarios. One of the most important capabilities for a robot building a representation of space and using that representation to localize and describe the environment is to reliably determine which observations come from the part of the environment that is known and represented by the model and which were not previously observed. This is a challenging task in case of purely appearance-based methods.

One of the streams of research at CVAP/CAS (http://www.cas.kth.se/) aims at developing a system for identifying places in an indoor environment (e.g. recognizing that this room is the kitchen that the system knows) and categorizing them (recognizing that this unknown room is a kitchen) based on visual data (images captured by a robot). In this project, the focus is on implementing, evaluating and developing methods for detecting novel rooms and room categories, i.e. discriminating between places known and unknown to the robot.

The student will work on extending a visual place classification system with the novelty detection capabilities. Starting with a baseline method based on the Kernel PCA algorithm, the student will implement and evaluate other proposed novelty detection methods and will adapt them to the problem of visual place classification. The system will be applied in the scenario considered in the [email protected]'10 (http://www.robotvision.info/) challenge and the method developed by the student will participate in the challenge and will be described in a paper submitted to the ImageCLEF proceedings published by Springer in the LNCS series.

In the considered scenario, the system will classify rooms and functional areas on the basis of image sequences, captured by a stereo camera mounted on a mobile robot within an office environment. The test sequence will be acquired within the same building but at a different floor than the training sequence. It will contain rooms of the same categorical type (corridor, office, bathroom) and it will also contain room categories not seen in the training sequence (meeting room, library). The system should be able to answer the question where are you? when presented with a test sequence imaging a room category seen during training, and it should be able to answer I do not know this category when presented with a new room category.

Requirements: background in computer science, mathematics, engineering or equivalent. Fundamental knowledge in pattern recognition and machine learning. Strong scientific programming skills (Matlab, scripting languages, C++ is a plus). Experience with Linux-based computing environments. The candidate is required to be ambitious and innovative.

Applications should be made as soon as possible. Potential applicants are invited to email a CV, study report, and a very brief email summarizing the personal interests and skills in:
- pattern recognition / machine learning
- robotics
- programming, computing environments, and operating systems

- Patric Jensfelt (http://www.csc.kth.se/~patric/)
- Andrzej Pronobis (http://www.pronobis.pro/)

Contact: Andrzej Pronobis (http://www.pronobis.pro/)


Informationen om uppsatsförslag är hämtad från Nationella Exjobb-poolen.