Use the Force

This example trains and uses a simple machine learning model using JSR-381.

Before getting started, the example was created using IntelliJ and JDK11 of GraalVM. You may need to adjust the set paths in the code to different relative locations. And using any other Java 11 JVM should do fine as well.

Training the model

The architecture of the model is small, it's based on that we used for MNIST, simply changed input and output layers. The training data is originally from Nick Bourdakos.

Run this Java file to train the model:

jsr381.starwars.TrainModel

Afterwards, it will export the trained model to deepnetts/trained_model.dnet

Using the trained model

You can use the already provided trained model or you can train your own and use that one. In case you'd like to use your own trained model, modify the path of importModel(...) in jsr381.starwars.UseModel

Run this Java file to use the model:

jsr381.starwars.UseModel

JavaFX application

To start the JavaFX application, run mvn javafx:run in the root of the project.

• You'd have to open the model first by going to menu: File -> Open Model...
• Select a DeepNetts trained model like from this project.
• Then open an image by going to the menu: File -> Open Image...
• Select a proper image and it will show up in the center of the application
• Then click on classify to get the classification results.