Merge pull request #9754 from HelgaShiryaeva/cnn-dl4j-java-example

CNN example with Deeplearning4j in Java

deeplearning4j/pom.xml

@@ -37,6 +37,16 @@
             <artifactId>deeplearning4j-nn</artifactId>
             <version>${dl4j.version}</version>
         </dependency>
+        <dependency>
+            <groupId>org.slf4j</groupId>
+            <artifactId>slf4j-api</artifactId>
+            <version>${slf4j.version}</version>
+        </dependency>
+        <dependency>
+            <groupId>org.slf4j</groupId>
+            <artifactId>slf4j-log4j12</artifactId>
+            <version>${slf4j.version}</version>
+        </dependency>
         <!-- https://mvnrepository.com/artifact/org.datavec/datavec-api -->
         <dependency>
             <groupId>org.datavec</groupId>
@@ -53,6 +63,7 @@
     <properties>
         <dl4j.version>0.9.1</dl4j.version> <!-- Latest non beta version -->
         <httpclient.version>4.3.5</httpclient.version>
+        <slf4j.version>1.7.5</slf4j.version>
     </properties>
 
 </project>

deeplearning4j/src/main/java/com/baeldung/deeplearning4j/cnn/CifarDataSetService.java

@@ -0,0 +1,47 @@
+package com.baeldung.deeplearning4j.cnn;
+
+import lombok.Getter;
+import org.deeplearning4j.datasets.iterator.impl.CifarDataSetIterator;
+import org.deeplearning4j.nn.conf.inputs.InputType;
+import org.nd4j.linalg.dataset.api.iterator.DataSetIterator;
+
+import java.util.List;
+
+@Getter
+class CifarDataSetService implements IDataSetService {
+
+    private final InputType inputType = InputType.convolutional(32, 32, 3);
+    private final int trainImagesNum = 512;
+    private final int testImagesNum = 128;
+    private final int trainBatch = 16;
+    private final int testBatch = 8;
+
+    private final CifarDataSetIterator trainIterator;
+
+    private final CifarDataSetIterator testIterator;
+
+    CifarDataSetService() {
+        trainIterator = new CifarDataSetIterator(trainBatch, trainImagesNum, true);
+        testIterator = new CifarDataSetIterator(testBatch, testImagesNum, false);
+    }
+
+    @Override
+    public DataSetIterator trainIterator() {
+        return trainIterator;
+    }
+
+    @Override
+    public DataSetIterator testIterator() {
+        return testIterator;
+    }
+
+    @Override
+    public InputType inputType() {
+        return inputType;
+    }
+
+    @Override
+    public List<String> labels() {
+        return trainIterator.getLabels();
+    }
+}

deeplearning4j/src/main/java/com/baeldung/deeplearning4j/cnn/CnnExample.java

@@ -0,0 +1,18 @@
+package com.baeldung.deeplearning4j.cnn;
+
+
+import lombok.extern.slf4j.Slf4j;
+import org.deeplearning4j.eval.Evaluation;
+
+@Slf4j
+class CnnExample {
+
+    public static void main(String... args) {
+        CnnModel network = new CnnModel(new CifarDataSetService(), new CnnModelProperties());
+
+        network.train();
+        Evaluation evaluation = network.evaluate();
+
+        log.info(evaluation.stats());
+    }
+}

deeplearning4j/src/main/java/com/baeldung/deeplearning4j/cnn/CnnModel.java

@@ -0,0 +1,119 @@
+package com.baeldung.deeplearning4j.cnn;
+
+import lombok.extern.slf4j.Slf4j;
+import org.deeplearning4j.eval.Evaluation;
+import org.deeplearning4j.nn.api.OptimizationAlgorithm;
+import org.deeplearning4j.nn.conf.MultiLayerConfiguration;
+import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
+import org.deeplearning4j.nn.conf.layers.ConvolutionLayer;
+import org.deeplearning4j.nn.conf.layers.OutputLayer;
+import org.deeplearning4j.nn.conf.layers.SubsamplingLayer;
+import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
+import org.deeplearning4j.nn.weights.WeightInit;
+import org.nd4j.linalg.activations.Activation;
+import org.nd4j.linalg.lossfunctions.LossFunctions;
+
+import java.util.stream.IntStream;
+
+@Slf4j
+class CnnModel {
+
+    private final IDataSetService dataSetService;
+
+    private final MultiLayerNetwork network;
+
+    private final CnnModelProperties properties;
+
+    CnnModel(IDataSetService dataSetService, CnnModelProperties properties) {
+
+        this.dataSetService = dataSetService;
+        this.properties = properties;
+
+        MultiLayerConfiguration configuration = new NeuralNetConfiguration.Builder()
+          .seed(1611)
+          .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT)
+          .learningRate(properties.getLearningRate())
+          .regularization(true)
+          .updater(properties.getOptimizer())
+          .list()
+          .layer(0, conv5x5())
+          .layer(1, pooling2x2Stride2())
+          .layer(2, conv3x3Stride1Padding2())
+          .layer(3, pooling2x2Stride1())
+          .layer(4, conv3x3Stride1Padding1())
+          .layer(5, pooling2x2Stride1())
+          .layer(6, dense())
+          .pretrain(false)
+          .backprop(true)
+          .setInputType(dataSetService.inputType())
+          .build();
+
+        network = new MultiLayerNetwork(configuration);
+    }
+
+    void train() {
+        network.init();
+        int epochsNum = properties.getEpochsNum();
+        IntStream.range(1, epochsNum + 1).forEach(epoch -> {
+            log.info("Epoch {} / {}", epoch, epochsNum);
+            network.fit(dataSetService.trainIterator());
+        });
+    }
+
+    Evaluation evaluate() {
+        return network.evaluate(dataSetService.testIterator());
+    }
+
+    private ConvolutionLayer conv5x5() {
+        return new ConvolutionLayer.Builder(5, 5)
+                .nIn(3)
+                .nOut(16)
+                .stride(1, 1)
+                .padding(1, 1)
+                .weightInit(WeightInit.XAVIER_UNIFORM)
+                .activation(Activation.RELU)
+                .build();
+    }
+
+    private SubsamplingLayer pooling2x2Stride2() {
+        return new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX)
+                .kernelSize(2, 2)
+                .stride(2, 2)
+                .build();
+    }
+
+    private ConvolutionLayer conv3x3Stride1Padding2() {
+        return new ConvolutionLayer.Builder(3, 3)
+                .nOut(32)
+                .stride(1, 1)
+                .padding(2, 2)
+                .weightInit(WeightInit.XAVIER_UNIFORM)
+                .activation(Activation.RELU)
+                .build();
+    }
+
+    private SubsamplingLayer pooling2x2Stride1() {
+        return new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX)
+                .kernelSize(2, 2)
+                .stride(1, 1)
+                .build();
+    }
+
+    private ConvolutionLayer conv3x3Stride1Padding1() {
+        return new ConvolutionLayer.Builder(3, 3)
+                .nOut(64)
+                .stride(1, 1)
+                .padding(1, 1)
+                .weightInit(WeightInit.XAVIER_UNIFORM)
+                .activation(Activation.RELU)
+                .build();
+    }
+
+    private OutputLayer dense() {
+        return new OutputLayer.Builder(LossFunctions.LossFunction.MEAN_SQUARED_LOGARITHMIC_ERROR)
+                .activation(Activation.SOFTMAX)
+                .weightInit(WeightInit.XAVIER_UNIFORM)
+                .nOut(dataSetService.labels().size() - 1)
+                .build();
+    }
+}

deeplearning4j/src/main/java/com/baeldung/deeplearning4j/cnn/CnnModelProperties.java

@@ -0,0 +1,13 @@
+package com.baeldung.deeplearning4j.cnn;
+
+import lombok.Value;
+import org.deeplearning4j.nn.conf.Updater;
+
+@Value
+class CnnModelProperties {
+    private final int epochsNum = 512;
+
+    private final double learningRate = 0.001;
+
+    private final Updater optimizer = Updater.ADAM;
+}

deeplearning4j/src/main/java/com/baeldung/deeplearning4j/cnn/IDataSetService.java

@@ -0,0 +1,16 @@
+package com.baeldung.deeplearning4j.cnn;
+
+import org.deeplearning4j.nn.conf.inputs.InputType;
+import org.nd4j.linalg.dataset.api.iterator.DataSetIterator;
+
+import java.util.List;
+
+interface IDataSetService {
+    DataSetIterator trainIterator();
+
+    DataSetIterator testIterator();
+
+    InputType inputType();
+
+    List<String> labels();
+}