dataset/
├── apples/
│   ├── mature/
│   │   ├── apple1.jpg
│   │   ├── apple2.jpg
│   │   └──...
│   └── defective/
│       ├── apple3.jpg
│       ├── apple4.jpg
│       └──...
├── bananas/
│   ├── mature/
│   │   ├── banana1.jpg
│   │   ├── banana2.jpg
│   │   └──...
│   └── unripe/
│       ├── banana3.jpg
│       ├── banana4.jpg
│       └──...
└──...

<dependency>
    <groupId>org.deeplearning4j</groupId>
    <artifactId>deeplearning4j-core</artifactId>
    <version>1.0.0-M1.4</version>
</dependency>
<dependency>
    <groupId>org.deeplearning4j</groupId>
    <artifactId>deeplearning4j-nn</artifactId>
    <version>1.0.0-M1.4</version>
</dependency>
<dependency>
    <groupId>org.deeplearning4j</groupId>
    <artifactId>deeplearning4j-ui</artifactId>
    <version>1.0.0-M1.4</version>
</dependency>
<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-web</artifactId>
</dependency>

import org.datavec.image.loader.NativeImageLoader;
import org.deeplearning4j.datasets.iterator.impl.ListDataSetIterator;
import org.nd4j.linalg.dataset.api.preprocessor.DataNormalization;
import org.nd4j.linalg.dataset.api.preprocessor.ImagePreProcessingScaler;
import java.io.File;
import java.util.ArrayList;
import java.util.List;
public class DataLoader {
    public static ListDataSetIterator loadData(String dataDirectory) {
        NativeImageLoader imageLoader = new NativeImageLoader(28, 28, 3);
        List<File> imageFiles = new ArrayList<>();
        // 遍歷數(shù)據(jù)集目錄，收集圖像文件
        File[] directories = new File(dataDirectory).listFiles(File::isDirectory);
        for (File directory : directories) {
            File[] files = directory.listFiles();
            for (File file : files) {
                imageFiles.add(file);
            }
        }
        // 創(chuàng)建圖像數(shù)據(jù)集
        float[][] images = new float[imageFiles.size()][28 * 28 * 3];
        int[] labels = new int[imageFiles.size()];
        for (int i = 0; i < imageFiles.size(); i++) {
            File imageFile = imageFiles.get(i);
            String label = imageFile.getParentFile().getName();
            labels[i] = Integer.parseInt(label);
            float[] image = imageLoader.asMatrix(imageFile);
            images[i] = image;
        }
        // 數(shù)據(jù)歸一化
        DataNormalization scaler = new ImagePreProcessingScaler(0, 1);
        scaler.fit(images);
        scaler.transform(images);
        // 創(chuàng)建數(shù)據(jù)集迭代器
        return new ListDataSetIterator(images, labels, 10);
    }
}

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.DenseLayer;
import org.deeplearning4j.nn.conf.layers.OutputLayer;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.deeplearning4j.nn.weights.WeightInit;
import org.nd4j.linalg.activations.Activation;
import org.nd4j.linalg.lossfunctions.LossFunctions;
public class FruitQualityDetector {
    public static MultiLayerNetwork buildModel() {
        MultiLayerConfiguration configuration = new NeuralNetConfiguration.Builder()
               .seed(123)
               .weightInit(WeightInit.XAVIER)
               .updater(org.deeplearning4j.nn.weights.WeightInit.XAVIER_UNIFORM)
               .list()
               .layer(0, new ConvolutionLayer.Builder(5, 5)
                       .nIn(3)
                       .stride(1, 1)
                       .nOut(16)
                       .activation(Activation.RELU)
                       .build())
               .layer(1, new ConvolutionLayer.Builder(3, 3)
                       .stride(1, 1)
                       .nOut(32)
                       .activation(Activation.RELU)
                       .build())
               .layer(2, new DenseLayer.Builder()
                       .nOut(512)
                       .activation(Activation.RELU)
                       .build())
               .layer(3, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD)
                       .nOut(2)
                       .activation(Activation.SOFTMAX)
                       .build())
               .build();
        return new MultiLayerNetwork(configuration);
    }
    public static void trainModel(MultiLayerNetwork model, ListDataSetIterator iterator, int epochs) {
        model.init();
        for (int i = 0; i < epochs; i++) {
            model.fit(iterator);
            iterator.reset();
        }
    }
}

import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.deeplearning4j.eval.Evaluation;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.dataset.api.preprocessor.DataNormalization;
import org.nd4j.linalg.dataset.api.preprocessor.ImagePreProcessingScaler;
import org.datavec.image.loader.NativeImageLoader;
import java.io.File;
public class ModelEvaluator {
    public static void evaluateModel(MultiLayerNetwork model, String testDataDirectory) {
        NativeImageLoader imageLoader = new NativeImageLoader(28, 28, 3);
        File[] testFiles = new File(testDataDirectory).listFiles();
        float[][] testImages = new float[testFiles.length][28 * 28 * 3];
        int[] testLabels = new int[testFiles.length];
        for (int i = 0; i < testFiles.length; i++) {
            File testFile = testFiles[i];
            String label = testFile.getParentFile().getName();
            testLabels[i] = Integer.parseInt(label);
            float[] image = imageLoader.asMatrix(testFile);
            testImages[i] = image;
        }
        DataNormalization scaler = new ImagePreProcessingScaler(0, 1);
        scaler.fit(testImages);
        scaler.transform(testImages);
        Evaluation evaluation = model.evaluate(new ListDataSetIterator(testImages, testLabels, 10));
        System.out.println(evaluation.stats());
    }
}

import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.deeplearning4j.datasets.iterator.impl.ListDataSetIterator;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Test;
import static org.junit.jupiter.api.Assertions.assertEquals;
public class FruitQualityDetectorTest {
    private String dataDirectory;
    private MultiLayerNetwork model;
    private ListDataSetIterator iterator;
    @BeforeEach
    public void setUp() {
        dataDirectory = "path/to/dataset";
        iterator = DataLoader.loadData(dataDirectory);
        model = FruitQualityDetector.buildModel();
    }
    @Test
    public void testBuildModel() {
        assertEquals(4, model.getLayers().length);
    }
    @Test
    public void testTrainModel() {
        FruitQualityDetector.trainModel(model, iterator, 10);
        assertEquals(true, model.conf().getLayers()[0].getParam("W").isInitialized());
    }
    @Test
    public void testEvaluateModel() {
        String testDataDirectory = "path/to/test/dataset";
        ModelEvaluator.evaluateModel(model, testDataDirectory);
        // 可以根據(jù)具體的評估指標(biāo)進行斷言
    }
}