Intelligent Mobile Projects with TensorFlow

Using the basic GAN model

In the runMNISTModel method, we first prepare a random input to the model:

void runMNISTModel() {    float[] floatValues = new float[BATCH_SIZE*100];    Random r = new Random();    for (int i=0; i<BATCH_SIZE; i++) {        for (int j=0; i<100; i++) {            double sample = r.nextGaussian();            floatValues[i] = (float)sample;        }    }

Then feed the input to the model, run the model and gets the output values, which are scaled grayscale values between 0.0 and 1.0, and convert them to integers in the range of 0 and 255:

    float[] outputValues = new float[BATCH_SIZE * 28 * 28];    AssetManager assetManager = getAssets();    mInferenceInterface = new TensorFlowInferenceInterface(assetManager, MODEL_FILE1); mInferenceInterface.feed(INPUT_NODE1, floatValues, ...