Computational Problem Solving (5 cr)

Evaluation scale

Approved/Rejected

Objective

- Understand and apply mathematical and computational methods for reconstructing movement and drawings from video footage.
- Analyze camera perspectives, distortions, and projections to infer spatial relationships.
- Design, implement, and evaluate simulations and visual explainers.
- Apply optimization techniques such as local search, gradient descent, and neural networks to improve model accuracy.
- Implement the solution using a programming language (eg. JavaScript or Python).

Content

The course is structured around a single challenge: reconstructing a pen’s path from footage of its movement in front of colored balls. Through this, students will explore:
- Trilateration and geometric localization techniques.
- Perspective analysis and size scaling based on visual input.
- Camera modeling (pinhole camera, lens distortion).
- Map projections (Azimuthal equidistant, Lambert equal-area).
- Simulations and visual debugging (using JavaScript and Three.js).
- Optimization strategies: local search, gradient descent, and genetic algorithms.
- Neural network applications in spatial estimation.
- Signal and image processing techniques for segmentation and motion analysis.
- Problem decomposition and algorithm design for real-world-inspired scenarios.

Location and time

Online

Materials

Original course notes and explainers by the instructor
Szeliski, R. Computer Vision: Algorithms and Applications (optional)

Teaching methods

The course is organized in 2 phases.

Phase 1 - The Competition (16.5.2025 - 15.7.2025)

Link: https://youtu.be/bZ8uSzZv0ew
In this phase, students self-study material they find online, and discuss with the teacher in the forum on Discord (decode-the-drawings channel):
https://discord.com/invite/gJFcF5XVn9

Students implement their own solutions and get feedback from the teacher. Depending on their progress, students get badges which translate to credits in this course.

Phase 2 - The Course (1.8.2025 - 15.12.2025)
The course ask for a number of homework assignments to be completed. Students can choose from a long list, up to 5 assignments to implement (1 assignment = 1 credit point).

Assessment criteria, approved/failed

The course is passed by earning a single badge during the competition, or doing one single homework assignment during the course.
The number of badges / homework assignments increase the number of credits (max 5 credits).