java深度学习之DJL卷积神经网络CNN

伪代码

代码使用3层卷积神经网络conv 和两层全连接Fully Connected Layer

// conv -> conv -> conv -> fc -> fc
        return new SequentialBlock()
                .add(Conv2d.builder()
                         //卷积核大小 8*8
                        .setKernelShape(new Shape(8, 8))
                        //步幅 4*4
                        .optStride(new Shape(4, 4))
                        .optPadding(new Shape(3, 3))
                        //卷积核层数
                        .setFilters(4).build())
                .add(Activation::relu)

                .add(Conv2d.builder()
                        .setKernelShape(new Shape(4, 4))
                        .optStride(new Shape(2, 2))
                        .setFilters(32).build())
                .add(Activation::relu)

                .add(Conv2d.builder()
                        .setKernelShape(new Shape(3, 3))
                        .optStride(new Shape(1, 1))
                        .setFilters(64).build())
                .add(Activation::relu)

                .add(Blocks.batchFlattenBlock())
                .add(Linear
                        .builder()
                        .setUnits(512).build())
                .add(Activation::relu)

                .add(Linear
                        .builder()
                        .setUnits(2).build());

DJL使用pytorch模型识别狗的种类

pom

  <dependency>
            <groupId>ai.djl.pytorch</groupId>
            <artifactId>pytorch-engine</artifactId>
            <version>0.16.0</version>
        </dependency>
        <dependency>
            <groupId>ai.djl.pytorch</groupId>
            <artifactId>pytorch-native-auto</artifactId>
            <version>1.9.1</version>
            <scope>runtime</scope>
        </dependency>
        <dependency>
            <groupId>ai.djl.pytorch</groupId>
            <artifactId>pytorch-jni</artifactId>
            <version>1.9.1-0.16.0</version>
            <scope>runtime</scope>
        </dependency>

java文件

package com.example.demo.util;

import ai.djl.MalformedModelException;
import ai.djl.Model;
import ai.djl.inference.Predictor;
import ai.djl.modality.Classifications;
import ai.djl.modality.cv.Image;
import ai.djl.modality.cv.ImageFactory;
import ai.djl.modality.cv.transform.CenterCrop;
import ai.djl.modality.cv.transform.Resize;
import ai.djl.modality.cv.transform.ToTensor;
import ai.djl.modality.cv.translator.ImageClassificationTranslator;
import ai.djl.repository.zoo.ModelNotFoundException;
import ai.djl.training.util.DownloadUtils;
import ai.djl.training.util.ProgressBar;
import ai.djl.translate.Pipeline;
import ai.djl.translate.TranslateException;
import ai.djl.translate.Translator;

import java.io.IOException;
import java.nio.file.Path;
import java.nio.file.Paths;

//狗狗种类预测
public class PyTorchLearn {
    public static void main(String[] args) throws IOException, TranslateException, MalformedModelException, ModelNotFoundException {
        //下载模型到本地
        DownloadUtils.download("https://djl-ai.s3.amazonaws.com/mlrepo/model/cv/image_classification/ai/djl/pytorch/resnet/0.0.1/traced_resnet18.pt.gz", "build/pytorch_models/resnet18/resnet18.pt", new ProgressBar());

        DownloadUtils.download("https://djl-ai.s3.amazonaws.com/mlrepo/model/cv/image_classification/ai/djl/pytorch/synset.txt", "build/pytorch_models/resnet18/synset.txt", new ProgressBar());
        Path modelDir = Paths.get("build/pytorch_models/resnet18");
        //加载模型
        Model model = Model.newInstance("resnet");
        model.load(modelDir, "resnet18");
//        先创建一个管道（每个图像要经过的预处理）
        Pipeline pipeline = new Pipeline();
        pipeline.add(new CenterCrop()).add(new Resize(224, 224)).add(new ToTensor());
//        然后创建转换器
        Translator<Image, Classifications> translator = ImageClassificationTranslator.builder()
                .setPipeline(pipeline)
                //载入所有标签进去
                .optSynsetArtifactName("synset.txt")
                //如果你的模型最后一层没有经过softmax就启用它
                .optApplySoftmax(true)
                //最终显示概率最高的3个
                .optTopK(3)
                .build();
        Image img = ImageFactory.getInstance().fromUrl("https://img1.baidu.com/it/u=281078762,293009126&fm=253&fmt=auto&app=138&f=JPEG?w=750&h=500");
        img.getWrappedImage();
//        执行推理
        Predictor<Image, Classifications> predictor = model.newPredictor(translator);
        Classifications classifications = predictor.predict(img);
        System.out.println(classifications);
    }
}