Java code with comments

package org.deeplearning4j.examples.modelimport.keras.basic;

import org.deeplearning4j.nn.graph.ComputationGraph;
import org.deeplearning4j.nn.modelimport.keras.KerasModelImport;
import org.deeplearning4j.nn.modelimport.keras.exceptions.InvalidKerasConfigurationException;
import org.deeplearning4j.nn.modelimport.keras.exceptions.UnsupportedKerasConfigurationException;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.ops.CustomOp;
import org.nd4j.linalg.api.ops.DynamicCustomOp;
import org.nd4j.linalg.factory.Nd4j;

import java.io.IOException;


public class TestImport {

    public static void main(String[] args) throws UnsupportedKerasConfigurationException, IOException, InvalidKerasConfigurationException {

        //Keras code included below..
        final String MODEL_PATH = "/Users/susaneraly/SKYMIND/de_conv.h5";
        INDArray in = Nd4j.linspace(0, 17, 18).reshape(1, 3, 3, 2).permute(0, 3, 1, 2);

        //Load the keras model
        ComputationGraph model = KerasModelImport.importKerasModelAndWeights(MODEL_PATH);
        INDArray modelOutput = model.output(in)[0];
        System.out.println(modelOutput.permute(0, 2, 3, 1));
        System.out.println("=======================");

        //This is the call to libnd4j that happens when model.output is called - I am just replicating it here manually
        INDArray weights = Nd4j.ones(2, 2, 2, 2);
        INDArray bias = Nd4j.zeros(1, 2);
        INDArray[] opInputs = new INDArray[]{in, weights, bias};
        INDArray[] opOutputs = new INDArray[]{Nd4j.ones(1, 2, 6, 6)};
        int[] argsA = new int[]{2, 2, 2, 2, 4, 4, 1, 1, 1, 0};
        CustomOp op = DynamicCustomOp.builder("deconv2d")
                .addInputs(opInputs)
                .addIntegerArguments(argsA)
                .addOutputs(opOutputs)
                .callInplace(false)
                .build();
        Nd4j.getExecutioner().exec(op);
        System.out.println(opOutputs[0].permute(0, 2, 3, 1));

        /*
        Both should give, but don't:
                [[[[ 1.  1.]
                   [ 1.  1.]
                   [ 5.  5.]
                   [ 5.  5.]
                   [ 9.  9.]
                   [ 9.  9.]]

                  [[ 1.  1.]
                   [ 1.  1.]
                   [ 5.  5.]
                   [ 5.  5.]
                   [ 9.  9.]
                   [ 9.  9.]]

                  [[13. 13.]
                   [13. 13.]
                   [17. 17.]
                   [17. 17.]
                   [21. 21.]
                   [21. 21.]]

                  [[13. 13.]
                   [13. 13.]
                   [17. 17.]
                   [17. 17.]
                   [21. 21.]
                   [21. 21.]]

                  [[25. 25.]
                   [25. 25.]
                   [29. 29.]
                   [29. 29.]
                   [33. 33.]
                   [33. 33.]]

                  [[25. 25.]
                   [25. 25.]
                   [29. 29.]
                   [29. 29.]
                   [33. 33.]
                   [33. 33.]]]]
        */
    }

}

keras code

from __future__ import print_function
import numpy as np
import keras
from keras.layers import Conv2DTranspose, Input
from keras.models import Model

input_shape=(3, 3, 2)
inputs = Input(shape=input_shape, name='encoder_input')
outputs = Conv2DTranspose(filters=2,kernel_size=(2,2),strides=2,padding='same')(inputs)

model = Model(inputs, outputs, name='decoder')
model.layers[1].set_weights([np.ones((2,2,2,2)),np.zeros((2,))])
a = np.arange(18).reshape(1,3,3,2)
aa = model.predict(a)
print(aa)
model.save("de_conv.h5")

/*
output of deconv is:
[[[[ 1.  1.]
   [ 1.  1.]
   [ 5.  5.]
   [ 5.  5.]
   [ 9.  9.]
   [ 9.  9.]]

  [[ 1.  1.]
   [ 1.  1.]
   [ 5.  5.]
   [ 5.  5.]
   [ 9.  9.]
   [ 9.  9.]]

  [[13. 13.]
   [13. 13.]
   [17. 17.]
   [17. 17.]
   [21. 21.]
   [21. 21.]]

  [[13. 13.]
   [13. 13.]
   [17. 17.]
   [17. 17.]
   [21. 21.]
   [21. 21.]]

  [[25. 25.]
   [25. 25.]
   [29. 29.]
   [29. 29.]
   [33. 33.]
   [33. 33.]]

  [[25. 25.]
   [25. 25.]
   [29. 29.]
   [29. 29.]
   [33. 33.]
   [33. 33.]]]]

*/