<![CDATA[In disaster scenarios, accurate and efficient human detection is essential to support timely search and rescue operations. This study explores the performance of deep learning models YOLOv8n for human detection using datasets before and after data augmentation. The evaluation focuses on key metrics including Precision, Recall, F1-score, inference time, and mean Average Precision (mAP). Experimental results indicate that the model trained on the original dataset achieves Precision (0.9623), Recall (0.9464), and F1-score (0.9357), highlighting better accuracy in minimizing false positives and false negatives. Conversely, the augmented dataset leads to improvements in mAP (95.8 vs. 94.5) and inference speed (8.2 ms vs. 9 ms), demonstrating increased robustness and efficiency. These findings suggest that while training on unaugmented data slightly better detection accuracy, data augmentation enhances the model's overall performance, speed and perform well to detect object in occluction scenario, making the YOLOv8n model more suitable for real-time usage in disaster response scenarios.]]>
Copyrights © 2025
