During the course the students learn to combine and apply basic results, algorithms, and methodology of 3D computer vision to solve vision problems in practical applications.
A short hands-on course about the basic results and methodology of 3D computer vision, focusing on geometric results and optimization, and their practical applications in, for example, augmented reality, motion capture, robotics, and vision-based interaction techniques.
Content:
Projective geometry: homogeneous coordinates, perspective projection, single view geometry, absolute pose problem (DLT & P3P), camera calibration, two view geometry, relative pose problem, triangulation. (4+4h)
Estimation: Feature detection & matching, random sample consensus, Levenberg-Marquardt optimization, automatic differentiation, large-scale sparse optimization (bundle adjustment), 3D reconstruction from multiple views or video. (4+4h)
Dense 3D: Dense stereo, depth cameras (e.g. Kinect), point clouds, iterative closest point algorithm. (2+2h)
Applications & open source libraries for computer vision (exercises & project)
Coursework exercises (75%) and a short computer vision project (25%), no exam.
Course notes, slides, and selected papers. The course covers short selected
parts of
This course can be included on the modules
Prerequisites:
The course focuses on solving practical problems rather than on theory. Basics of programming (C++, Python, or Octave/Matlab), linear algebra (working with small matrices e.g. in computer graphics), calculus (derivatives), and probability (counting combinations, normal distribution) suffice.
Language of Instruction:
Lectures in Finnish or English depending on enrolled students. The course material is in English.