Research on Optimizing the YOLOV8n Detection Model for Fatigue Detection in Smart Lamps

In recent years, with the fusion of artificial intelligence and smart furniture, health-based smart furniture has become popular, among which smart desk lamps have attracted much attention due to their significant impact on vision health. Studies have shown that LED light significantly affects the retina, and the number of near-sighted people worldwide is expected to reach 4.76 billion by 2025. Long-term incorrect use of light is a major cause of myopia, and eye-protecting desk lamps can effectively reduce this hazard. This study combines the YOLOV8n neural network model with a smart desk lamp system and optimizes its design to automatically adjust light attributes according to the user's needs and state, providing a comfortable lighting environment, alleviating eye fatigue, and realizing facial fatigue recognition. The optimization of the YOLOV8n(You Only Look Once version 8Nano) model introduces an attentional mechanism and improves the backbone network, enabling it to perform well in fatigue expression recognition. The hardware part uses ESP32-CMA for image acquisition and networking, combined with Arduino microcontroller to control external devices. Experiments include feasibility, ablation, and comparison experiments between different models. Results show that the improved model with CoordAttention and EfficientViT modules outperforms the unoptimized model in recognition accuracy, recall, and mAP0.5 metrics. Compared with other common target detection models, the proposed model has clear advantages in recognition accuracy and recall. This study brings important innovations in smart home technology, capturing face features more accurately, improving user experience, and supporting vision health.