DeliveredBodytrak · 2017 — 2019
Person-down detection
Neural-network fall detection on accelerometer + gyroscope data; 92% accuracy.
My first engineering role: a neural-network person-down / fall-detection system analysing accelerometer and gyroscope data for real-time detection in high-risk environments — 92% accuracy, with MLflow for tracking.
Key features
- Neural-network fall detection on IMU data
- 92% detection accuracy
Built with — how & why
- PythonModel and data pipeline.
- pandasAccelerometer / gyroscope feature prep.
- TensorFlowThe fall-detection neural network.
