parthmanekar25/Driver_Fatigue_Detection_Analysis ? reverse-engineered prompt

Build me a real time driver fatigue detection project that works well on a Raspberry Pi or other low power device.

I want it to watch a driver through a camera and detect early signs of fatigue like eye closure and yawning, then trigger an alert when the driver looks tired. Please use a hybrid approach, so it combines a lightweight image based model with facial landmark based checks like eye aspect ratio, mouth aspect ratio, and head pose, since that should cut down on false alarms.

It should include training and inference code, simple preprocessing for eye and mouth images, and a clear pipeline from camera input to fatigue decision to alert output. Make it efficient enough for live use, and include a way to test accuracy, latency, FPS, and false alarms. If anything is missing, fill in sensible defaults and look up current docs online if you need to.