Part-II : Converting(Saving) Keras Model as a single HDF5 (.h5) file

4 Part blog series : Documenting the learnings of my experimentation on Object Detection using Keras, Tensorflow

Photo by Tamanna Rumee on Unsplash

If you have read the part-I blog ,you can re-collect that we had used wpod-net pre-trained model with separate model weight (*.h5) and model structure (*.json) files. Both these files were loaded into the python program in-order to detect number plates.

In this blog post we will learn how to convert (save) keras model weight and model structure files into a single Keras *.h5 file

Introduction

A Keras model consists of multiple components:

• The architecture or configuration, which specifies what layers the model contain, and how they’re connected, typically in jsonor YAML format.
• A set of weights values (the “state of the model”), in .h5 format
• An optimiser (defined by compiling the model).

The Keras API makes it possible to save all of these pieces to disk at once. Saving everything into a single archive in the TensorFlow SavedModel format (or in the older Keras H5 format).

Implementation

If you like to follow along, refer to 2_convert_keras_model_to_single_file.ipynb jupyter notebook from my repository. Here are the implementation steps, the following program loads up keras model weight and model configuration files and saves them into a single model file by using model.save("models/wpod-net_all_io_one.h5") line.

• Once keras model is saved as a single file, list the models directory. Here you will find the file wpod_net_all_in_one.h5
• Finally, let’s use this newly created keras model file and try to detect license plate.

• You should get results like below

Summary

• In this blog we have learned how to save keras model file *.h5 and the configuration file *.json as one file *.h5 keras model file
• We have also verified that, this new all-in-one model file is able to detect number plate from the image, as we wanted
• In the next blog , we will convert this keras model all-in-one file into tensorflow (*.pb) frozengraph model file and its text representation