Skip to content

Latest commit

 

History

History
81 lines (53 loc) · 3.63 KB

README.md

File metadata and controls

81 lines (53 loc) · 3.63 KB
Raw

Converting a Model file to CoreML

The docker image comes out of the box with a couple of example models and convertion scripts. To build and run:

./build-and-run.sh

When the container is running, navigate to notebooks/notebook/Convert Karas Model.ipynb and execute the script blocks (by clicking the Run button on top of the notebook), this will:

  • step through the differrent blocks in the notebook (note: wait for the script blocks to finish executing before executing the next block!)
  • download the Food 101 Keras model and labels (to /models)
  • convert the Food 101 Keras model to a CoreML model
  • test the converted model
  • save the CoreML model (to the root)

Customization

While the example notebook that converts the Food 101 Keras model to a CoreML model works out of the box, you might find it more convenient to directly download the model files when building the Docker image. You can download them in the Dockerfile by adding these lines after the workspace directory is created:

# Download example Food101 Keras model weights and labels
RUN wget -O /workspace/models/keras-food101-model.hdf5 https://s3.amazonaws.com/stratospark/food-101/model4b.10-0.68.hdf5
RUN wget -O /workspace/models/keras-food101-model-labels.txt https://github.com/stratospark/food-101-mobile/raw/43598fdc08500683bbc04f877ae069c38c8ac4c3/model_export/labels.txt

Build the Docker Image

To build a docker image using a DockerFile, issue the following command:

docker build -t convert-coreml .

Run docker container

Run the docker container using the image you just built:

docker run --rm -it -p 8888:8888 -v $(pwd)/notebook:/workspace/notebook convert-coreml

—-rm: when finished using this container, remove it and clean it up

—it: open an interactive shell when the container starts

-p local-port:container-port: forward our local port to the container’s port

-v local-dir:container-dir: mount a local volume in the container

Launch Jupyter

Browse to https://localhost:8888 and check out the notebooks.

Note

The keras-food101-model is unfortunately only compatible with keras 1.2.2 and Tensorflow 1.0.1, so they have been pinned to these versions in requirements.txt. Also, the model prediction step is only available when running on macOS (so not inside a Docker container) so it is expected to fail. The convertion to a Core ML model works fine though.

Additional Resources

License

See the accompanying LICENSE file for more information.

Copyright 2018 Jeroen Wesbeek

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.