This project implements a real-time motion classification system on an STM32 microcontroller using a pretrained and quantized machine learning model generated with Edge Impulse (https://www.edgeimpulse.com/)
The system captures motion data from a sensor (accelerometer), performs preprocessing, and runs on-device inference to classify motion patterns.
- Data collection using motion sensor
- Dataset labeling
- Model training in Edge Impulse
- Model quantization (int8)
- Deployment to STM32
- Real-time inference on-device