(open the full-size APo image in a new tab)APo: Real-Time Elderly Activity Monitoring
- Role
- Machine Learning Engineer
Keeping track of an older adult’s activity usually takes someone watching, in person or on camera. APo recognizes activities from a camera feed in real time on a Raspberry Pi and shows the latest detected activity in a mobile app.
My contribution
Developed and optimized a motion pre-filtering and MobileNet-based activity recognition pipeline.
- Python
- TensorFlow
- OpenCV
- MediaPipe
- MobileNet
- Raspberry Pi
Reported evaluation results
- Accuracy
- 98.83%
- Mean latency
- 25.25 ms
- Average CPU usage
- 71.39%Reduced from 79.41%








