Examples with FeedForwardToCnnPreProcessor org.deeplearning4j.nn.conf.preprocessor.FeedForwardToCnnPreProcessor used on opensource projects

Example 1 with FeedForwardToCnnPreProcessor

use of org.deeplearning4j.nn.conf.preprocessor.FeedForwardToCnnPreProcessor in project deeplearning4j by deeplearning4j.

the class TestRenders method testHistogramComputationGraph.

@Test
public void testHistogramComputationGraph() throws Exception {
    ComputationGraphConfiguration conf = new NeuralNetConfiguration.Builder().optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT).graphBuilder().addInputs("input").addLayer("cnn1", new ConvolutionLayer.Builder(2, 2).stride(2, 2).nIn(1).nOut(3).build(), "input").addLayer("cnn2", new ConvolutionLayer.Builder(4, 4).stride(2, 2).padding(1, 1).nIn(1).nOut(3).build(), "input").addLayer("max1", new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX).kernelSize(2, 2).build(), "cnn1", "cnn2").addLayer("output", new OutputLayer.Builder().nIn(7 * 7 * 6).nOut(10).build(), "max1").setOutputs("output").inputPreProcessor("cnn1", new FeedForwardToCnnPreProcessor(28, 28, 1)).inputPreProcessor("cnn2", new FeedForwardToCnnPreProcessor(28, 28, 1)).inputPreProcessor("output", new CnnToFeedForwardPreProcessor(7, 7, 6)).pretrain(false).backprop(true).build();
    ComputationGraph graph = new ComputationGraph(conf);
    graph.init();
    graph.setListeners(new HistogramIterationListener(1), new ScoreIterationListener(1));
    DataSetIterator mnist = new MnistDataSetIterator(32, 640, false, true, false, 12345);
    graph.fit(mnist);
}

Example 2 with FeedForwardToCnnPreProcessor

use of org.deeplearning4j.nn.conf.preprocessor.FeedForwardToCnnPreProcessor in project deeplearning4j by deeplearning4j.

the class ComputationGraphConfigurationTest method testJSONBasic2.

@Test
public void testJSONBasic2() {
    ComputationGraphConfiguration conf = new NeuralNetConfiguration.Builder().optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT).graphBuilder().addInputs("input").addLayer("cnn1", new ConvolutionLayer.Builder(2, 2).stride(2, 2).nIn(1).nOut(5).build(), "input").addLayer("cnn2", new ConvolutionLayer.Builder(2, 2).stride(2, 2).nIn(1).nOut(5).build(), "input").addLayer("max1", new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX).kernelSize(2, 2).build(), "cnn1", "cnn2").addLayer("dnn1", new DenseLayer.Builder().nOut(7).build(), "max1").addLayer("max2", new SubsamplingLayer.Builder().build(), "max1").addLayer("output", new OutputLayer.Builder().nIn(7).nOut(10).build(), "dnn1", "max2").setOutputs("output").inputPreProcessor("cnn1", new FeedForwardToCnnPreProcessor(32, 32, 3)).inputPreProcessor("cnn2", new FeedForwardToCnnPreProcessor(32, 32, 3)).inputPreProcessor("dnn1", new CnnToFeedForwardPreProcessor(8, 8, 5)).pretrain(false).backprop(true).build();
    String json = conf.toJson();
    ComputationGraphConfiguration conf2 = ComputationGraphConfiguration.fromJson(json);
    assertEquals(json, conf2.toJson());
    assertEquals(conf, conf2);
}

Example 3 with FeedForwardToCnnPreProcessor

use of org.deeplearning4j.nn.conf.preprocessor.FeedForwardToCnnPreProcessor in project deeplearning4j by deeplearning4j.

the class ConvolutionLayerSetupTest method complete.

public MultiLayerConfiguration.Builder complete() {
    final int numRows = 28;
    final int numColumns = 28;
    int nChannels = 1;
    int outputNum = 10;
    int iterations = 10;
    int seed = 123;
    MultiLayerConfiguration.Builder builder = new NeuralNetConfiguration.Builder().seed(seed).iterations(iterations).optimizationAlgo(OptimizationAlgorithm.LINE_GRADIENT_DESCENT).list().layer(0, new org.deeplearning4j.nn.conf.layers.ConvolutionLayer.Builder(new int[] { 10, 10 }, new int[] { 2, 2 }).nIn(nChannels).nOut(6).build()).layer(1, new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX, new int[] { 2, 2 }).weightInit(WeightInit.XAVIER).activation(Activation.RELU).build()).layer(2, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD).nIn(//216
    5 * 5 * 1 * 6).nOut(outputNum).weightInit(WeightInit.XAVIER).activation(Activation.SOFTMAX).build()).inputPreProcessor(0, new FeedForwardToCnnPreProcessor(numRows, numColumns, nChannels)).inputPreProcessor(2, new CnnToFeedForwardPreProcessor(5, 5, 6)).backprop(true).pretrain(false);
    return builder;
}

Example 4 with FeedForwardToCnnPreProcessor

use of org.deeplearning4j.nn.conf.preprocessor.FeedForwardToCnnPreProcessor in project deeplearning4j by deeplearning4j.

the class TestComputationGraphNetwork method testCnnFlatInputType1.

@Test
public void testCnnFlatInputType1() {
    //First: check conv input type. Expect: no preprocessor, nIn set appropriately
    ComputationGraphConfiguration conf = new NeuralNetConfiguration.Builder().graphBuilder().addInputs("in").setInputTypes(InputType.convolutional(10, 8, 3)).addLayer("layer", new ConvolutionLayer.Builder().kernelSize(2, 2).padding(0, 0).stride(1, 1).build(), "in").addLayer("out", new OutputLayer.Builder().nOut(10).build(), "layer").setOutputs("out").pretrain(false).backprop(true).build();
    LayerVertex lv = (LayerVertex) conf.getVertices().get("layer");
    FeedForwardLayer l = ((FeedForwardLayer) (lv).getLayerConf().getLayer());
    assertEquals(3, l.getNIn());
    assertNull(lv.getPreProcessor());
    //Check the equivalent config, but with flat conv data input instead
    //In this case, the only difference should be the addition of a preprocessor
    //First: check conv input type. Expect: no preprocessor, nIn set appropriately
    conf = new NeuralNetConfiguration.Builder().graphBuilder().addInputs("in").setInputTypes(InputType.convolutionalFlat(10, 8, 3)).addLayer("layer", new ConvolutionLayer.Builder().kernelSize(2, 2).padding(0, 0).stride(1, 1).build(), "in").addLayer("out", new OutputLayer.Builder().nOut(10).build(), "layer").setOutputs("out").pretrain(false).backprop(true).build();
    lv = (LayerVertex) conf.getVertices().get("layer");
    l = ((FeedForwardLayer) (lv).getLayerConf().getLayer());
    assertEquals(3, l.getNIn());
    assertNotNull(lv.getPreProcessor());
    InputPreProcessor preProcessor = lv.getPreProcessor();
    assertTrue(preProcessor instanceof FeedForwardToCnnPreProcessor);
    FeedForwardToCnnPreProcessor preproc = (FeedForwardToCnnPreProcessor) preProcessor;
    assertEquals(10, preproc.getInputHeight());
    assertEquals(8, preproc.getInputWidth());
    assertEquals(3, preproc.getNumChannels());
    //Finally, check configuration with a subsampling layer
    conf = new NeuralNetConfiguration.Builder().graphBuilder().addInputs("in").setInputTypes(InputType.convolutionalFlat(10, 8, 3)).addLayer("l0", new SubsamplingLayer.Builder().kernelSize(2, 2).stride(1, 1).padding(0, 0).build(), "in").addLayer("layer", new ConvolutionLayer.Builder().kernelSize(2, 2).padding(0, 0).stride(1, 1).build(), "l0").addLayer("out", new OutputLayer.Builder().nOut(10).build(), "layer").setOutputs("out").pretrain(false).backprop(true).build();
    //Check subsampling layer:
    lv = (LayerVertex) conf.getVertices().get("l0");
    SubsamplingLayer sl = ((SubsamplingLayer) (lv).getLayerConf().getLayer());
    assertNotNull(lv.getPreProcessor());
    preProcessor = lv.getPreProcessor();
    assertTrue(preProcessor instanceof FeedForwardToCnnPreProcessor);
    preproc = (FeedForwardToCnnPreProcessor) preProcessor;
    assertEquals(10, preproc.getInputHeight());
    assertEquals(8, preproc.getInputWidth());
    assertEquals(3, preproc.getNumChannels());
    //Check dense layer
    lv = (LayerVertex) conf.getVertices().get("layer");
    l = ((FeedForwardLayer) (lv).getLayerConf().getLayer());
    assertEquals(3, l.getNIn());
    assertNull(lv.getPreProcessor());
}