Research on Real-Time Human Fall Detection Method Based on YOLOv5-Lite

Fall detection is an important application with various applications in fields such as healthcare and safety monitoring. This paper presents a fall detection method based on the lightweight object detection model YOLOv5-Lite. The proposed method uses a custom dataset for training and optimizing the model to achieve precise detection of human fall states in real-time scenarios. Accuracy, recall rate, mAP, and FPS are used to evaluate the model's performance. Experimental results show that when the IoU is 0.5, the accuracy of the model is 92.1%, with a detection accuracy of 94.0% and a recall rate of 85.6% for fall states. Regarding the running speed, the model runs at 3.12ms FPS when using a camera with a
$640\times 480$
size as input on an RTX3060 (12G) graphics card. This study has practical and promotional value and provides technical support and reference for human fall detection in practical scenarios such as elderly care facilities and public areas.