CIFAR-10 image classfier using Keras

This design is very similar to the TensorFlow version, but uses Keras exclusively.

The example demonstrates the following:

  • Simple download of CIFAR-10 dataset.
  • Scaling of the image data, one-hot encoding of the labels.
  • How to define a CNN using the Keras sequential model
  • How to set up Keras callbacks for Tensorboard logging, early stopping and checkpoint saving.
  • How to compile, train, evaluate the model.
  • How to make predictions with the trained model.
  • How to save the Keras model as HDF5 and JSON formats.
  • How to freeze the graph to a protobuf file from Keras.

The code and a Jupyter notebook is available on my Github page.

If you have any questions or comments about this design, please email me at