This work presents the design, implementation, and validation of a real-time system for the automatic detection of face mask usage, using modern computer vision techniques on low-cost embedded hardware. The solution was based on the SSD-MobileNetV2 architecture, trained with environment-specific images and executed locally on an NVIDIA Jetson Nano board, which enabled a balance between accuracy, inference speed, and cost. The system was evaluated under seven controlled scenarios that considered variations in distance, movement, lateral orientation, use of eyeglasses, and presence of facial hair. It achieved accuracy rates above 95% at distances shorter than one meter and detection latencies below 0.6 s. These results demonstrated the robustness and practical applicability of the proposed approach in real-world contexts.