TY  - EJOU
AU  - Calderón, Luis Sánchez 
AU  - Burneo, David Valverde 
AU  - Orrala, Walter Hurtares 

TI  - AI-Driven Object Detection Framework for Live Load Monitoring and Structural Optimization
T2  - Structural Durability \& Health Monitoring

PY  - 
VL  - 
IS  - 
SN  - 1930-2991

AB  - Accurate characterization of live load histories remains critical for structural safety and efficient design; however, traditional codes often overestimate in-service loads. This study introduced an AI-driven framework integrating YOLOv8 object detection and DeepFace gender classification with continuous video surveillance to monitor live loads in academic buildings. Gender classification used local anthropometric data (77 kg males, 61 kg females) for precise load estimation, with privacy ensured via local processing and anonymized metadata only. Observed peaks were substantially below Eurocode and IBC provisions, confirming code conservatism. Uncertainty propagation from detector errors (recall 0.57, ±0.02 Kn/m<sup>2</sup>) minimally impacted projections. These findings demonstrate the potential of computer vision for data-driven structural optimization and sustainable design.
KW  - Live loads; computer vision; Convolutional Neural Networks (CNN); video surveillance; structural engineering; YOLOv8; DeepFace; occupancy monitoring

DO  - 10.32604/sdhm.2026.077137