Deep Learning for Construction Image Identification: A Comparative Analysis

Construction site monitoring is vital for project management, safety compliance, and progress tracking. The advent of deep learning has revolutionized computer vision capabilities, enabling automated identification and classification of construction images. This paper presents a novel multi-scale feature fusion network (MS-FFNet) for construction image identification and provides a comprehensive comparison with state-of-the-art models including ResNet-50, EfficientNet-B3, and Vision Transformer (ViT). This paper evaluates these models on a diverse construction image dataset comprising 15,000 images across 12 categories of construction activities and elements. Experimental results demonstrate that the proposed MS-FFNet achieves 94.7% accuracy, outperforming baseline models while maintaining computational efficiency. Paper provides detailed analysis of model performance across different construction categories, lighting conditions, and occlusion levels. The proposed model shows particular strength in distinguishing between visually similar construction elements and maintaining performance in challenging environmental conditions.