Notes

All of the projects proposed below are related to computer vision and pattern recognition. However, they also include important components related to algorithm design, database design, artificial intelligence, machine learning, data mining, software engineering, graph theory, user interfaces, mathematics, statistics, etc.
I am not expecting you to have a computer vision or pattern recognition background but you should be motivated and enthusiastic about learning new topics.
I expect the students to work consistently throughout two semesters and to have periodic meetings with me to discuss their progress.
These projects can be extended to M.S./Ph.D. research in the future.
Please contact me if you are interested in any of the projects below or would like to propose your own project.

Projects

Object recognition for content-based image retrieval
Content-based image retrieval (CBIR) refers to the retrieval of images according to their content, rather than through standard keyword-based techniques. Early work on CBIR used color and texture properties of images to model their similarities. However, users want to search for images according to higher-level concepts such as scene classes (indoor/outdoor, city/landscape) and the objects present in the scene (cars, people, animals, buildings). Object recognition is one of the fundamental but unsolved problems in computer vision. This project involves the design and implementation of a system that uses low-level cues from color, texture, shape, edges, corners, etc. to build higher level object models for semantic classification and retrieval in image databases. You should be good at C/C++ or Java programming and have interest in computer vision, pattern recognition, machine learning, graph theory and user interfaces.
Event and object models for video classification and retrieval
Efficient categorization, browsing and retrieval of videos in large databases require detection of people, objects and events in these videos. Both visual properties such as color, texture, shape and motion, and audio and speech data provide important information about the content of a video. This project consists of the design and implementation of a system that integrates multi-modal information for classification and content-based retrieval from video archives. You should be good at C/C++ or Java programming and have interest in computer vision, pattern recognition, machine learning and data mining.
Video object segmentation and tracking
Video object segmentation involves extraction of objects of interest from video sequences and video object tracking involves identification of these objects' trajectories and relations both in time and space. Important applications of video object segmentation and tracking include robot and autonomous vehicle navigation, detection of special events in surveillance videos, medical therapy such as monitoring the behaviors of physical therapy patients, and intelligent transportation systems that model pedestrian and highway traffic. Other applications include MPEG-4 and MPEG-7 that are the successors to MPEG-1 and MPEG-2 that made interactive video on VCDs and DVDs possible. This project involves the design and implementation of automatic methods for identifying objects of interest and tracking them in video sequences. You should be good at C/C++ or Java programming and have interest in computer vision, computer graphics, graph theory and user interfaces.
Multi-scale and multi-temporal image data and feature server
Advances in computational power and imaging technology is constantly increasing the quantity, variability and availability of images from different sources. Two important applications where there is a great need for automatic intelligent analysis of images is remote sensing and medical imaging. For satellite images, automatic categorization of land covers, detection of different scene structures and analysis of changes in these structures require intelligent storage and processing of images taken at different resolutions (multi-scale) and different times (multi-temporal). For example, detection of different parks or residential areas can be done in low resolution but identification of individual tree groups or buildings require higher resolution images. In addition, analysis of the effects of changes in urbanization on environment needs images of the same area to be taken periodically. Similarly, identification of cells and their nuclei need different resolution images of a tissue sample. This project consists of the design and implementation of an intelligent system that supports efficient representation, storage, retrieval and processing of multi-scale and multi-temporal images. Design of algorithms for the example applications mentioned above will also be studied after a successful implementation of the database component. You should be good at C++ and Java programming and have interest in computer vision, pattern recognition, database design and user interfaces.

Students

IRIS - Information Retrieval as Integration of Image and Semantics
- Pınar Kavak
- Ahmet Kızıl
- Emre Küçükayvaz
Remote Sensing Image Classification and Retrieval
- Özer Akargün
- Kazım Umut Aşkın
- Başak Bilgi
- Abdülkadir Tokgöz

Schedule

Please check the list of important deadlines for the 2005-2006 Academic Year.

Announcements

Notes

Projects

Students

Schedule

Related Links