Towards Open-World Understanding of Construction Tools and Equipment
Bachelor's Thesis — Sofia Storozhenko (2026)
Student: Sofia Storozhenko
Type: Bachelor’s Thesis in Informatics
Submitted: March 16, 2026
Supervision / advising team: Panagiotis Petropoulakis, André Borrmann, Alois C. Knoll
This thesis proposes a two-stage framework for open-world understanding of construction tools and equipment. A real-time YOLOv12 detector first localizes tools, after which a Vision-Language Model provides semantic interpretation and hazard reasoning beyond rigid predefined class labels.
The project also develops a synthetic data-generation pipeline for construction-floor scenes and evaluates the combination of fast object detection with multimodal reasoning for safer robotic perception in dynamic construction environments.