Are Open-Vocabulary Models Ready for Detection of MEP Elements on Construction Sites

2025-01-16

AI Summary

This study addresses the need for accurate small-object detection of MEP (mechanical, electrical, and plumbing) components by ground robots on construction sites. Method: We systematically compare open-vocabulary vision-language models—specifically CLIP-based zero-shot detection—with lightweight closed-set detectors (YOLOv5/v8 fine-tuned on domain-specific data). Our evaluation is grounded in a manually annotated image dataset captured by ground robots in real construction environments—the first empirical benchmark of open-vocabulary models for professional small-object detection. Results: Domain-adapted lightweight detectors achieve significantly higher mAP than open-vocabulary approaches (+23.6%), demonstrating that task-specific optimization remains indispensable under edge-computing constraints. Crucially, we identify domain adaptation as a key performance bottleneck for open-vocabulary models in specialized industrial settings. Our work establishes a reproducible benchmark and practical deployment pathway for small-object recognition in construction automation.

Technology Category

• Computer Vision: Object Detection & Categorization
• Intelligent Robots: Learning & Optimization for ROB
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

The construction industry has long explored robotics and computer vision, yet their deployment on construction sites remains very limited. These technologies have the potential to revolutionize traditional workflows by enhancing accuracy, efficiency, and safety in construction management. Ground robots equipped with advanced vision systems could automate tasks such as monitoring mechanical, electrical, and plumbing (MEP) systems. The present research evaluates the applicability of open-vocabulary vision-language models compared to fine-tuned, lightweight, closed-set object detectors for detecting MEP components using a mobile ground robotic platform. A dataset collected with cameras mounted on a ground robot was manually annotated and analyzed to compare model performance. The results demonstrate that, despite the versatility of vision-language models, fine-tuned lightweight models still largely outperform them in specialized environments and for domain-specific tasks.

Problem

• Image Recognition
• Construction Sites
• MEP Components

Innovation

• Specialized Image Recognition Models
• MEP Components Identification
• Site-specific Accuracy Improvement