- Students have knowledge of advanced computer vision and pattern recognition techniques and are able to apply them.
- Students can read and understand scientific papers on advanced computer vision and pattern recognition and are able to implement (part of) the methods described to verify results reported in the papers.
This course is a follow-up of the ‘image processing and computer vision’ course and as such a very good preparation for students who want to do their master's assignment in biometrics (Master specialisation Computer Vision and Biometrics), medical imaging, or robotics. The course addresses advanced methods for image segmentation, object tracking and motion analysis in images, 3D computer vision, and 3D reconstruction from e.g. stereo images and statistical pattern recognition for image analysis, e.g. face recognition.
The course consists of four main subjects:
- Basics of statistical pattern recognition: Classification, Detection, Learning, Principle Component Analysis (PCA), Linear Discriminant Analysis (LDA), Parameter estimation.
- Advanced image segmentation techniques: Mean shift, Snakes, Level-set, Active Shape Model (ASM), Watershed
- Motion analysis: Tracking in 3D, Kalman filter, Optic flow
- 3D computer vision and 3D image analysis: Fundamentals, Stereo, 3D vision
- Topics from Deep Learning (DL) and Convolutional Neural Networks (CNN)
Form of the course
1. The students have to prepare for the lectures and present the basic methods they studied and their application to the projects.
The course is project based, meaning the students actively acquire knowledge of the above mentioned subjects by applying them and investigating their performance. The course consists of a number of lectures where students present the acquired and applied knowledge followed by discussion with their fellow students and the teachers. The projects are carried out in groups of 2-4 students (individual projects are also possible). The students can choose the subjects for the projects from a predefined list that changes every year. Examples of projects in previous years are: lip reading from video, people tracking across multiple cameras and 3D reconstruction using 2D video, face recognition from video. For each project a number of basic methods must be studied and their application investigated. During the final part of the course, the aim is to realize a working system resulting in a demonstration and a report in the form of a publication.
2. The students have to write a report of about 6 pages in the form of a scientific paper
3. At the end of the course there will be closing event where all students are present and each of them presents the results in a short presentation and if possible, a demonstration.
4. To get a grade, a student must fulfill the first 3 conditions.
The final mark is based on the report.