Course overview
Over the last 40 years, researchers in artificial intelligence have endeavoured to develop computers with the capacity to "see" the world around them. This course aims to convey the nature of some of the fundamental problems in vision, and to explain a variety of techniques used to overcome them. Vision is a rapidly evolving area of computer science, and new and emerging approaches to these problems are discussed along with more "classical" techniques. Various vision problems are considered, including: feature detection in images, e.g. edge detection, and the accumulation of edge data to form lines; recovery of 3D shape from images, e.g. the use of a stereo image pair to derive 3D surface information; forming image mosaics; video surveillance techniques, e.g. tracking objects in video; motion detection in video images, e.g. counting number of moving objects in a video; recognising and classifying objects in images, e.g. searching a video for a particular object. Several assignments will be given to enable the student to gain practical experience in tackling some of these problems.
Course learning outcomes
- Describe the scope of challenges and applications addressed by computer vision
- Demonstrate and experiment with image filtering techniques
- Make use of geometric camera models and multiple view geometry
- Undertake video analysis problems such as tracking and structure from motion
- Explain the application of neural networks to computer vision
- Analyse cognitive tasks including image classification, recognition and detection
- Conduct computer vision experiments and report results systematically