Structural defects such as cracks, corrosion, honeycombing, and dampness pose significant risks to infrastructure safety. This study developed and fine-tuned a YOLOv8n AI model, which improves the detection accuracy of these common defects. By using enhanced data and optimized settings, the model provides an efficient solution to support structural health monitoring, reducing risks to personnel inspecting high-rise structures. The fine-tuning results reveal significant gains: honeycomb detection achieved a 0.6% increase in box precision, a 0.4% rise in box recall, a 1.7% improvement in mean average precision reaching 0.817 for boxes and 0.792 for masks, and measure of harmonic mean of precision and recall of 0.810 (Box) and 0.800 (Mask); crack detection saw a 0.5% improvement in box recall, a 0.6% rise in mean average precision (0.860 for boxes, 0.870 for masks), and measure of harmonic mean of precision and recall of 0.790 (Box) and 0.810 (Mask); damp-strain detection experienced a 3.1% increase in box recall, a 2.1% boost in mean average precision (0.859 for boxes, 0.824 for masks), and measure of harmonic mean of precision and recall of 0.840 (Box) and 0.830 (Mask); and corrosion detection benefited from a 2.7% gain in box precision, a 0.5% improvement in box recall, a 1.0% increase in mean average precision (0.811 for boxes, 0.798 for masks), and measure of harmonic mean of precision and recall of 0.750 for both box and mask.