Ming Chu
Healthcare IT and Machine Learning Researcher
Bozeman, Montana
đ LinkedIn
by Ming Chu
It is late at night. Your eyelids grow heavy and your phone buzzes. For a few seconds your attention drifts, but the car does not stop moving.
In those moments, seconds decide lives.
The goal of our research was simple. Build an AI assistant that sees distraction or fatigue in real time and alerts the driver before it is too late.
This project began as my graduate research and became the focus of my 3-Minute Thesis presentation.
In three minutes I shared how computer vision and deep learning can detect distraction and drowsiness inside the car.
The research was later published in the Journal of Big Data (Springer Open):
âComprehensive Study of Driver Behavior Monitoring Systems Using Computer Vision and Machine Learning Techniquesâ
The system reached 99.1% accuracy on benchmarked datasets under controlled experimental conditions. It was a strong step toward safer roads.
Think of the AI system as two teammates working together.
Technically these are CNNs (Convolutional Neural Networks) and BiLSTMs (Bidirectional Long Short-Term Memory): CNNs and BiLSTMs are proven for spatial and temporal patterns. Vision Transformers extend this by learning global context across frames. Together they can see the quick gaze shift, hands off the wheel, or the collapsing body posture.
If placed in vehicles, this system could:
This is not surveillance of the driver. No identity tracking or behavioral profiling is performed. It is assistance for the road.
The next step is training Vision Transformers (ViT) on large, real-world driving datasets.
This can bring three improvements:
This project grew my skills in:
The greatest lesson was this. Technology is worth building only when it serves people like a caring passenger. Yet even the best system cannot watch over us perfectly. There is One who never sleeps and never looks away. He guards our steps, and in Him we find true safety.