Computer vision Shipped · 2024
Driver Fatigue Detection
Real-time computer vision, from sensor to alert
Local project
Computer vision 2024
System architecture
- 01
Sensor
Real-time video feed
- 02
Preprocessing
OpenCV — frame and face extraction
- 03
Inference
PyTorch model — drowsiness sign recognition
- 04
Alert
Hardware integration, fired live
A real-time fatigue detection system wiring a video feed, a PyTorch model trained to recognise drowsiness signs, and hardware integration that fires the alert. The whole pipeline — sensor to action — runs live.
What was built
- PyTorch model recognising drowsiness signals
- Real-time video processing with OpenCV
- Hardware-software integration for alert triggering
- Demonstrated live at a hackathon
The real challenge
Running a model in real time on constrained hardware changes everything: latency, preprocessing, and robustness to real-world conditions matter as much as raw accuracy.
Next project
Concrete Crack Segmentation