← All work

03 / Real-time machine learning

Multimodal Driver State

A multimodal driver-monitoring pipeline combining face mesh, facial analysis, eye aspect ratio, head pose, and audio signals into interpretable risk and intention alerts.

Problem

Driver state is a multimodal signal, not a single facial expression.

Approach

Combine face mesh, facial analysis, eye aspect ratio, head pose, and audio signals.

Evidence

Public repository and an explicit real-time performance target.

Stack

Python · OpenCV · MediaPipe · TensorFlow · DeepFace