Researchers at Edith Cowan University (ECU) have developed an AI model that can detect whether a driver is drunk, tired, or angry, all from a single video feed of their face. No breathalyzer, no blood test, just your face.

The model, named Jack of Many Faces, can identify blood alcohol concentration with nearly 90% accuracy and drowsiness with 95% accuracy. It can even classify the level of intoxication, sorting drivers into sober, moderate, or severely impaired categories.

Can your expressions prove you are a dangerous driver?

Beyond drunk driving, the researchers also track facial expressions, like anger. The reason is straightforward: aggressive emotional states can be just as dangerous as intoxication behind the wheel.

According to Dr. Syed Zulqarnain Gilani from ECU’s Centre of AI and Machine Learning, high levels of fatigue can mimic the impairment of being drunk, and anger can lead to road rage. By identifying all three simultaneously, the system gets a much clearer picture of driver safety.

Can it work in the dark?

The researchers also tackled a practical problem. Standard cameras struggle in low-light conditions. Their second model, BiFuseNet, combines regular video with infrared footage to improve detection in low ambient lighting conditions or at night. 

ECU PhD candidate Abdullah Tariq, who led the research, said, “Our rational was to develop a fully automated framework for estimating blood alcohol concentration by using RGB and IR video stream,” further adding that “Previously this was done manually by looking at pupil dilation and eye closing ratios, but these methods may not perform well with certain factors, such as varying lighting conditions.

It works by automatically tracking facial dynamics, such as eye blinking, subtle facial movements, and other cues that reveal a driver’s physical state.

Why does this matter?

Traditional methods like breathalyzers are accurate but they require active cooperation from the driver. They also demand significant human resources present on the ground to administer the tests. This technology works passively and continuously, making it far more practical for real-world use.

Drink driving is one of the biggest contributing factors in crashes, accounting for roughly 30% of road accidents globally. If this technology makes it into vehicles, the consequences for road safety could be enormous.