deeplearning4j训练MNIST数据集以及验证

训练模型官方示例

MNIST数据下载地址: http://github.com/myleott/mnist_png/raw/master/mnist_png.tar.gz

GitHub示例地址: https://github.com/deeplearning4j/deeplearning4j-examples/blob/master/dl4j-examples/src/main/java/org/deeplearning4j/examples/quickstart/modeling/convolution/LeNetMNISTReLu.java

/*******************************************************************************
 *
 *
 *
 * This program and the accompanying materials are made available under the
 * terms of the Apache License, Version 2.0 which is available at
 * https://www.apache.org/licenses/LICENSE-2.0.
 *  See the NOTICE file distributed with this work for additional
 *  information regarding copyright ownership.
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
 * License for the specific language governing permissions and limitations
 * under the License.
 *
 * SPDX-License-Identifier: Apache-2.0
 ******************************************************************************/

package org.deeplearning4j.examples.quickstart.modeling.convolution;

import org.datavec.api.io.labels.ParentPathLabelGenerator;
import org.datavec.api.split.FileSplit;
import org.datavec.image.loader.NativeImageLoader;
import org.datavec.image.recordreader.ImageRecordReader;
import org.deeplearning4j.datasets.datavec.RecordReaderDataSetIterator;
import org.deeplearning4j.nn.conf.MultiLayerConfiguration;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.inputs.InputType;
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.conf.layers.SubsamplingLayer;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.deeplearning4j.nn.weights.WeightInit;
import org.deeplearning4j.optimize.listeners.ScoreIterationListener;
import org.deeplearning4j.examples.utils.DataUtilities;
import org.deeplearning4j.util.ModelSerializer;
import org.nd4j.evaluation.classification.Evaluation;
import org.nd4j.linalg.activations.Activation;
import org.nd4j.linalg.dataset.api.iterator.DataSetIterator;
import org.nd4j.linalg.dataset.api.preprocessor.DataNormalization;
import org.nd4j.linalg.dataset.api.preprocessor.ImagePreProcessingScaler;
import org.nd4j.linalg.learning.config.Nesterovs;
import org.nd4j.linalg.lossfunctions.LossFunctions;
import org.nd4j.linalg.schedule.MapSchedule;
import org.nd4j.linalg.schedule.ScheduleType;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import java.io.File;
import java.util.HashMap;
import java.util.Map;
import java.util.Random;

/**
 * Implementation of LeNet-5 for handwritten digits image classification on MNIST dataset (99% accuracy)
 * <a href="http://yann.lecun.com/exdb/publis/pdf/lecun-01a.pdf">[LeCun et al., 1998. Gradient based learning applied to document recognition]</a>
 * Some minor changes are made to the architecture like using ReLU and identity activation instead of
 * sigmoid/tanh, max pooling instead of avg pooling and softmax output layer.
 * <p>
 * This example will download 15 Mb of data on the first run.
 *
 * @author hanlon
 * @author agibsonccc
 * @author fvaleri
 * @author dariuszzbyrad
 */
public class LeNetMNISTReLu {
    private static final Logger LOGGER = LoggerFactory.getLogger(LeNetMNISTReLu.class);
    //    private static final String BASE_PATH = System.getProperty("java.io.tmpdir") + "/mnist";
    private static final String BASE_PATH = "D:\\Documents\\Downloads\\mnist_png";
    private static final String DATA_URL = "http://github.com/myleott/mnist_png/raw/master/mnist_png.tar.gz";

    public static void main(String[] args) throws Exception {
        // 图片高度
        int height = 28;    // height of the picture in px
        // 图片宽度
        int width = 28;     // width of the picture in px
        // 通道 1 表示 黑白
        int channels = 1;   // single channel for grayscale images
        // 可能出现的结果数量 0-9 10个数字
        int outputNum = 10; // 10 digits classification
        // 批处理数量
        int batchSize = 54; // number of samples that will be propagated through the network in each iteration
        // 迭代次数
        int nEpochs = 1;    // number of training epochs
        // 随机数生成器
        int seed = 1234;    // number used to initialize a pseudorandom number generator.
        Random randNumGen = new Random(seed);

        LOGGER.info("Data load...");
        if (!new File(BASE_PATH + "/mnist_png").exists()) {

            LOGGER.debug("Data downloaded from {}", DATA_URL);
            String localFilePath = BASE_PATH + "/mnist_png.tar.gz";
            if (DataUtilities.downloadFile(DATA_URL, localFilePath)) {
                DataUtilities.extractTarGz(localFilePath, BASE_PATH);
            }
        }

        LOGGER.info("Data vectorization...");
        // vectorization of train data
        File trainData = new File(BASE_PATH + "/mnist_png/training");
        FileSplit trainSplit = new FileSplit(trainData, NativeImageLoader.ALLOWED_FORMATS, randNumGen);
        ParentPathLabelGenerator labelMaker = new ParentPathLabelGenerator(); // use parent directory name as the image label
        ImageRecordReader trainRR = new ImageRecordReader(height, width, channels, labelMaker);
        trainRR.initialize(trainSplit);
        // MNIST中的数据
        DataSetIterator trainIter = new RecordReaderDataSetIterator(trainRR, batchSize, 1, outputNum);

        // pixel values from 0-255 to 0-1 (min-max scaling)
        DataNormalization imageScaler = new ImagePreProcessingScaler();
        imageScaler.fit(trainIter);
        trainIter.setPreProcessor(imageScaler);

        // vectorization of test data
        File testData = new File(BASE_PATH + "/mnist_png/testing");
        FileSplit testSplit = new FileSplit(testData, NativeImageLoader.ALLOWED_FORMATS, randNumGen);
        ImageRecordReader testRR = new ImageRecordReader(height, width, channels, labelMaker);
        testRR.initialize(testSplit);
        DataSetIterator testIter = new RecordReaderDataSetIterator(testRR, batchSize, 1, outputNum);
        testIter.setPreProcessor(imageScaler); // same normalization for better results

        LOGGER.info("Network configuration and training...");
        // reduce the learning rate as the number of training epochs increases
        // iteration #, learning rate
        Map<Integer, Double> learningRateSchedule = new HashMap<>();
        learningRateSchedule.put(0, 0.06);
        learningRateSchedule.put(200, 0.05);
        learningRateSchedule.put(600, 0.028);
        learningRateSchedule.put(800, 0.0060);
        learningRateSchedule.put(1000, 0.001);

        MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
                .seed(seed)
                .l2(0.0005) // ridge regression value
                .updater(new Nesterovs(new MapSchedule(ScheduleType.ITERATION, learningRateSchedule)))
                .weightInit(WeightInit.XAVIER)
                .list()
                .layer(new ConvolutionLayer.Builder(5, 5)
                        .nIn(channels)
                        .stride(1, 1)
                        .nOut(20)
                        .activation(Activation.IDENTITY)
                        .build())
                .layer(new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX)
                        .kernelSize(2, 2)
                        .stride(2, 2)
                        .build())
                .layer(new ConvolutionLayer.Builder(5, 5)
                        .stride(1, 1) // nIn need not specified in later layers
                        .nOut(50)
                        .activation(Activation.IDENTITY)
                        .build())
                .layer(new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX)
                        .kernelSize(2, 2)
                        .stride(2, 2)
                        .build())
                .layer(new DenseLayer.Builder().activation(Activation.RELU)
                        .nOut(500)
                        .build())
                .layer(new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD)
                        .nOut(outputNum)
                        .activation(Activation.SOFTMAX)
                        .build())
                .setInputType(InputType.convolutionalFlat(height, width, channels)) // InputType.convolutional for normal image
                .build();

        MultiLayerNetwork net = new MultiLayerNetwork(conf);
        net.init();
        net.setListeners(new ScoreIterationListener(10));
        LOGGER.info("Total num of params: {}", net.numParams());

        // evaluation while training (the score should go down)
        for (int i = 0; i < nEpochs; i++) {
            net.fit(trainIter);
            LOGGER.info("Completed epoch {}", i);
            Evaluation eval = net.evaluate(testIter);
            LOGGER.info(eval.stats());

            trainIter.reset();
            testIter.reset();
        }

        File ministModelPath = new File(BASE_PATH + "/minist-model.zip");
        ModelSerializer.writeModel(net, ministModelPath, true);
        LOGGER.info("The MINIST model has been saved in {}", ministModelPath.getPath());
    }
}

验证模型

package org.deeplearning4j.examples.quickstart.modeling.convolution;

import org.datavec.image.loader.NativeImageLoader;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.dataset.api.preprocessor.DataNormalization;
import org.nd4j.linalg.dataset.api.preprocessor.ImagePreProcessingScaler;

import java.io.File;
import java.io.IOException;

/**
 * @description:
 * @author: Mr.Fang
 * @create: 2023-07-14 15:06
 **/

public class VerifyMNSIT {
    public static void main(String[] args) throws IOException {

        // 加载训练好的模型
        File modelFile = new File("D:\\Documents\\Downloads\\mnist_png\\minist-model.zip");
        MultiLayerNetwork model = MultiLayerNetwork.load(modelFile, true);

        // 加载待验证的图像
        File imageFile = new File("D:\\Documents\\Downloads\\mnist_png\\mnist_png\\testing\\8\\1717.png");
        NativeImageLoader loader = new NativeImageLoader(28, 28, 1);
        INDArray image = loader.asMatrix(imageFile);
        DataNormalization scaler = new ImagePreProcessingScaler(0, 1);
        scaler.transform(image);

        // 对图像进行预测
        INDArray output = model.output(image);
        int predictedLabel = output.argMax().getInt();
        // 在这行代码中，`output.argMax()`用于找到`output`中具有最大值的索引。`output`是一个包含模型的输出概率的NDArray对象。对于MNIST模型，输出是一个长度为10的向量，表示数字0到9的概率分布。
        //
        //`.argMax()`方法返回具有最大值的索引。例如，如果`output`的值为[0.1, 0.3, 0.2, 0.05, 0.25, 0.05, 0.05, 0.1, 0.05, 0.05]，则`.argMax()`将返回索引1，因为在位置1处的值0.3是最大的。
        //
        //最后，`.getInt()`方法将获取`.argMax()`的结果并将其转换为一个整数，表示预测的标签。在这个例子中，`predictedLabel`将包含模型预测的数字标签。
        //
        //简而言之，这行代码的作用是找到输出中概率最高的数字标签，以进行预测。
        System.out.println("Predicted label: " + predictedLabel);
    }
}

输出结果

o.n.l.f.Nd4jBackend - Loaded [CpuBackend] backend
o.n.n.NativeOpsHolder - Number of threads used for linear algebra: 6
o.n.l.c.n.CpuNDArrayFactory - Binary level Generic x86 optimization level AVX/AVX2
o.n.n.Nd4jBlas - Number of threads used for OpenMP BLAS: 6
o.n.l.a.o.e.DefaultOpExecutioner - Backend used: [CPU]; OS: [Windows 10]
o.n.l.a.o.e.DefaultOpExecutioner - Cores: [12]; Memory: [4.0GB];
o.n.l.a.o.e.DefaultOpExecutioner - Blas vendor: [OPENBLAS]
o.n.l.c.n.CpuBackend - Backend build information:
 GCC: "12.1.0"
STD version: 201103L
DEFAULT_ENGINE: samediff::ENGINE_CPU
HAVE_FLATBUFFERS
HAVE_OPENBLAS
o.d.n.m.MultiLayerNetwork - Starting MultiLayerNetwork with WorkspaceModes set to [training: ENABLED; inference: ENABLED], cacheMode set to [NONE]
Predicted label: 8