A flutter app that lets you experiment with a MNIST (28x28) tf-keras model in real-time!
Works over websockets, using the amazing Starlette web framework.
- Save model using
model.save(mnist_model.h5)
- Download model file to
client/server/mnist_model.h5
. - Deploy python server by running the following commands in the
server/
directory (on a publically accessible server) -
pip install -e .
uvicorn --port <server port> server:app
- Change
serverUrl
accordingly. - In the
client/
directory,flutter run
.
Thanks to this medium article for the hand drawing canvas code.