DINO-YOLO: Self-Supervised Pre-training for Data-Efficient Object Detection in Civil Engineering Applications

TL;DR

DINO-YOLO is a hybrid self-supervised object detection architecture that significantly improves accuracy in civil engineering applications with limited data, while maintaining real-time inference speeds.

Contribution

The paper introduces DINO-YOLO, combining YOLOv12 with DINOv3 transformers, and demonstrates its effectiveness in data-efficient detection for civil engineering tasks.

Findings

Achieves up to 88.6% improvement in detection accuracy on KITTI dataset.

Maintains real-time inference at 30-47 FPS despite added complexity.

Optimal performance with medium-scale architectures and dual integration points.

Abstract

Object detection in civil engineering applications is constrained by limited annotated data in specialized domains. We introduce DINO-YOLO, a hybrid architecture combining YOLOv12 with DINOv3 self-supervised vision transformers for data-efficient detection. DINOv3 features are strategically integrated at two locations: input preprocessing (P0) and mid-backbone enhancement (P3). Experimental validation demonstrates substantial improvements: Tunnel Segment Crack detection (648 images) achieves 12.4% improvement, Construction PPE (1K images) gains 13.7%, and KITTI (7K images) shows 88.6% improvement, while maintaining real-time inference (30-47 FPS). Systematic ablation across five YOLO scales and nine DINOv3 variants reveals that Medium-scale architectures achieve optimal performance with DualP0P3 integration (55.77% mAP@0.5), while Small-scale requires Triple Integration (53.63%). The 2-4x inference overhead (21-33ms versus 8-16ms baseline) remains acceptable for field deployment on NVIDIA RTX 5090. DINO-YOLO establishes state-of-the-art performance for civil engineering datasets (<10K images) while preserving computational efficiency, providing practical solutions for construction safety monitoring and infrastructure inspection in data-constrained environments.