Doktorandské kolokvium KAI - Matej Mok (4.5.2026)
v pondelok 4.5.2026 o 13:10 hod. v miestnosti I 9
Prednášajúci: Matej Mok
Názov: Hybrid 3D Computer Vision with Limited Data
Termín: 4.5.2026, 13:10 hod., I/9
Abstrakt:
3D computer vision aims to reconstruct scene geometry from images and use it to extract useful information like detecting objects, localization, and camera pose estimation. In practice, many of these methods require large amounts of annotated data, which are often difficult or expensive to obtain. This work investigates hybrid approaches that combine geometric methods with machine learning in data-constrained settings. We focus on the development of line-based feature detection and evaluate the usefulness of such features in downstream tasks, including object pose detection and estimation of relative pose of cameras. In parallel, we explore the generation of synthetic training data using computer graphics and assessing its usefulness for training vision models. Part of this research is the development of digital twins of 3D capturing devices, enabling the creation of realistic synthetic datasets and supporting novel learning paradigms in virtual environments.