Skip to content
This repository has been archived by the owner on Aug 12, 2024. It is now read-only.

Latest commit

 

History

History
100 lines (75 loc) · 3.2 KB

README.md

File metadata and controls

100 lines (75 loc) · 3.2 KB

Online Training on Raspberry Pi using TensorFlow

This example performs online training of a gesture detector on a Raspberry Pi, using live sensor readings from a BBC Micro:bit.

  1. Collect data from Micro:bit
  2. Train initial model using Google Colab, save a model checkpoint
  3. Load model checkpoint on a Raspberry Pi
  4. Continue training on the Raspberry Pi, using live sensor data from the Micro:bit

Micro:bit Programming

Using the Python Micro:bit Editor, flash a Micro:bit with microbit/device_code.py

Initial data collection

In this section, you will collect data for your own Micro:bit gesture.

  1. Find the serial port path associated with Micro:bit

    a. With the Micro:bit disconnected:

    On MacOS, use `ls /dev/cu.*`

On Windows, open Device Manager and expand "Ports (COM & LPT)"

    b. Connect the Micro:bit:

    On MacOS, use `ls /dev/cu.*`, you should see a path that looks similar to `/dev/cu.usbmodemXXXX`

On Windows, a new COMXX node should appear under "Ports (COM & LPT)"

  2. Edit acquire_data.py, update comport to the path found in step 1:

    On MacOS:

    comport = '/dev/cu.usbmodem144102' # Example MacOS path

    On Windows:

    comport = 'COM3' # Example Windows path

  3. Setup the data collection script

    conda activate diec
conda install pyserial-asyncio

  4. Run the data collection script.

    python acquire_data.py

    As the script is running:

    • press button A to perform the gesture (by moving the Micro:bit)
    • release button A when your gesture completes
    • repeat about 10 times to gather enough data

    Data will be saved to data.csv

Training

  1. Open notebook from Google Colab: edge_online_learning.ipynb
  2. Make a copy of the notebook
  3. Upload data.csv to your Colab storage
  4. Run each cell of the notebook, step by step
  5. Download *.pkl and *.h5 from Colab storage to your laptop

Online Learning

  1. Deploy the *.pkl and *.h5 files to the Raspberry Pi. You can use WinSCP (on Windows) or scp (on MacOS) to transfer files to the Raspberry Pi. You should copy the files to this folder:
~/diec/day4/rpi
  1. Connect the Micro:bit device to the Raspberry Pi 3, check its serial path:
ls /dev/ttyA*

Note down the serial path, e.g. /dev/ttyACM0

Important: the Micro:bit must be connected before the docker container is launched, in order for the container to find the serial device.

  1. From the Raspberry Pi 3, launch docker container
cd ~/diec/day4/docker
sh ./launch_docker.sh
  1. From the docker container on the Raspberry Pi, verify that the serial device can be seen:
ls /dev/ttyA*
  1. From the docker container on the Raspberry Pi, incrementally update the model in intervals of N timesteps:
cd day4/rpi
python3 incremental_train.py /dev/ttyACM0 --update_interval=2

(Substitute /dev/ttyACM0 with the path from step 2)

Troubleshooting

For troubleshooting training issues, the incremental_train.py script can also be run on the laptop with the Micro:bit connected. Use the appropriate serial path.