Optimization of AI Methods for Object Detection and Classification in Electrical Installation Drawings

In the electrical installation industry, digital tools support the planning, execution, and quality assurance of construction projects. However, the identification of installation objects, such as symbols in drawings and descriptions in project documents, is still largely performed manually.

Recent advances in computer vision, including SAM3 (Segment Anything Model 3) for image segmentation and YOLO (You Only Look Once) for object detection and classification, offer opportunities to automate this process. While YOLO is effective for object classification, it requires extensive manual annotation of training data and may not achieve the desired accuracy. This thesis investigates whether combining segmentation and object classification can reduce annotation effort while improving detection accuracy.

The project aims to develop and evaluate a prototype that integrates segmentation and classification for identifying electrical installation objects in digital drawings. The results will be assessed in terms of processing time and classification performance, with the additional goal of publishing the findings in a scientific paper.

The thesis will be carried out in collaboration with MBiz and is suitable for one or two students in Electrical Engineering and/or Computer Engineering. Programming skills and an interest in machine learning are required.