Enhancing Victim Detection In Disaster Scenarios A Yolov7 And Yolov8 Performance Study

In the aftermath of natural disasters like earthquakes, rapid identification of victims is vital for efficient rescue operations. The ability to locate victims accurately and quickly amidst debris and challenging conditions can make a critical difference in saving lives. This study evaluates the performance of two advanced object detection models, YOLOv7 and YOLOv8, for victim detection in disaster scenarios. Both models were trained on a specialized dataset that simulates post-disaster environments, incorporating diverse and realistic challenges such as occlusions, varying lighting conditions, and complex backgrounds.

The experimental results demonstrated an accuracy of 58% for YOLOv7 and a significantly improved accuracy of 81% for YOLOv8, showcasing the latter's superior capability in detecting human bodies among debris. Additionally, YOLOv8 outperformed YOLOv7 in terms of precision, recall, and detection speed, making it better suited for real-time applications. YOLOv7, while less accurate, demonstrated a faster inference time, which could be advantageous in scenarios requiring rapid initial assessments.

This comparative analysis not only highlights the advancements of YOLOv8 but also sheds light on the strengths and limitations of YOLOv7, particularly in handling occluded objects and detecting victims in cluttered environments. The findings underscore the potential of leveraging YOLOv8 in real-time disaster response systems, where quick and reliable victim identification is crucial. Future work could focus on enhancing the robustness of these models by incorporating data from diverse disaster scenarios, integrating additional sensor modalities such as infrared imaging or thermal cameras, and further optimizing model architectures to balance speed and accuracy.

These advancements can pave the way for more effective deployment of AI-powered tools in emergency rescue operations, improving outcomes and saving more lives during critical times.

Keywords: victim detection, disaster response, YOLOv7, YOLOv8, emergency rescue, deep learning models, accuracy comparison, real-time detection, precision, recall, robustness.