Dynamic Vision: Tracking Methods in Computer Vision
This seminar course aims to provide a comprehensive introduction to the topic of tracking and motion estimation methods in computer vision, exploring key areas such as optical flow, feature tracking, single- and multi-object tracking. We focus on both foundational approaches and cutting-edge developments, offering an interdisciplinary perspective that merges theoretical understanding with practical application. One of the distinguishing features of this seminar is the integration of expert guest speakers, who are leaders in both academia and industry. These sessions give students the opportunity to engage with professionals shaping the future of tracking technologies. The course is structured in a topic-based format, with each major area covered over a two-week span. Sessions include an introduction to fundamental principles, presented by the course organizers, followed by a guest lecture on recent advances. This balanced approach allows students to progressively build their knowledge while gaining insight into the latest innovations.
Content
- Optical Flow
- Feature Tracking
- Point Tracking
- Single-Object Tracking
- Multi-Object Tracking
- Dynamic 3D Vision (e.g. Gaussian Splatting)
Course Duration
15 October 2024 – 11 February 2025
Schedule
Mondays 10:00-11:30am
Learning Outcomes
As an interdisciplinary seminar, it provides diverse insights into the field of tracking methods in computer vision. The course covers the overlap between deep learning, motion estimation, and object tracking techniques. Upon successful completion, participants will have gained the following:
- a brief introduction to deep learning and its applications in computer vision
- an overview of motion estimation and object tracking methods, including optical flow, feature, and point tracking
- familiarity with common challenges and difficulties encountered in tracking approaches
- practice in effective science presentation, by presenting a recent topic
