uTensor enables motion recognition on microcontrollers. The model is trained with a modified Activity of Daily dataset recognizing 5 classes:
The project is also a reference implementation of sequential data processing with Mbed and uTensor.
For sensor setup, please refer to Train/HMP_Dataset/MANUAL.txt. The grove sensor is place flat on the back of user's right hand, with the connector socket oriented furthest away from the wrist.
- Mbed F413ZH board
- Grove Sheild
- Grove 3D digital accelerometer
- Recommend cloud9 environment
$ mbed import https://github.com/uTensor/ADL_demo $ cd ADL_demo $ mbed compile -m DISCO_F413ZH -t GCC_ARM --profile=uTensor/build_profile/release.json
- Ensure the Grove sensor is connected
- Locate the binary path from the terminal output, and flash it onto the board
For Training Instruction, please see Train/README.md