java-tensorflow-samples

Java sample codes on how to load tensorflow pretrained model file and predict based on these pretrained model files

Usage

Image Classification using Cifar10

Below show the demo codes of the Cifar10ImageClassifier which loads the cnn_cifar10.pb tensorflow model file, and uses it to do image classification:

package com.github.chen0040.tensorflow.classifiers.demo;

import com.github.chen0040.tensorflow.classifiers.utils.ResourceUtils;
import com.github.chen0040.tensorflow.classifiers.cifar10.Cifar10ImageClassifier;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import java.awt.image.BufferedImage;
import java.io.IOException;
import java.io.InputStream;

public class Cifar10ImageClassifierDemo {

    private static final Logger logger = LoggerFactory.getLogger(Cifar10ImageClassifierDemo.class);

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


        InputStream inputStream = ResourceUtils.getInputStream("tf_models/cnn_cifar10.pb");
        Cifar10ImageClassifier classifier = new Cifar10ImageClassifier();
        classifier.load_model(inputStream);

        String[] image_names = new String[] {
                "airplane1",
                "airplane2",
                "airplane3",
                "automobile1",
                "automobile2",
                "automobile3",
                "bird1",
                "bird2",
                "bird3",
                "cat1",
                "cat2",
                "cat3"
        };

        for(String image_name :image_names) {
            String image_path = "images/cifar10/" + image_name + ".png";
            BufferedImage img = ResourceUtils.getImage(image_path);
            String predicted_label = classifier.predict_image(img);
            logger.info("predicted class for {}: {}", image_name, predicted_label);
        }
    }
}

Image Classification using Inception

Below show the demo codes of the InceptionImageClassifier which loads the tensorflow_inception_graph.pb tensorflow model file, and uses it to do image classification:

package com.github.chen0040.tensorflow.classifiers.demo;

import com.github.chen0040.tensorflow.classifiers.inception.InceptionImageClassifier;
import com.github.chen0040.tensorflow.classifiers.utils.ResourceUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import java.awt.image.BufferedImage;
import java.io.IOException;

public class InceptionImageClassifierDemo {

    private static final Logger logger = LoggerFactory.getLogger(InceptionImageClassifierDemo.class);

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


        InceptionImageClassifier classifier = new InceptionImageClassifier();
        classifier.load_model(ResourceUtils.getInputStream("tf_models/tensorflow_inception_graph.pb"));
        classifier.load_labels(ResourceUtils.getInputStream("tf_models/imagenet_comp_graph_label_strings.txt"));

        String[] image_names = new String[] {
                "tiger",
                "lion"
        };

        for(String image_name :image_names) {
            String image_path = "images/inception/" + image_name + ".jpg";
            BufferedImage img = ResourceUtils.getImage(image_path);
            String predicted_label = classifier.predict_image(img);
            logger.info("predicted class for {}: {}", image_name, predicted_label);
        }
    }
}

Sentiment Analysis using 1D CNN

Below show the demo codes of the CnnSentimentClassifier which loads the wordvec_cnn.pb tensorflow model file, and uses it to do sentiment analysis:

import com.github.chen0040.tensorflow.classifiers.sentiment.models.CnnSentimentClassifier;
import com.github.chen0040.tensorflow.classifiers.sentiment.utils.ResourceUtils;

import java.io.IOException;
import java.util.List;

public class CnnSentimentClassifierDemo {
    public static void main(String[] args) throws IOException {
        CnnSentimentClassifier classifier = new CnnSentimentClassifier();
        classifier.load_model(ResourceUtils.getInputStream("tf_models/wordvec_cnn.pb"));
        classifier.load_vocab(ResourceUtils.getInputStream("tf_models/wordvec_cnn.csv"));

        List<String> lines = ResourceUtils.getLines("data/umich-sentiment-train.txt");
        for(String line : lines){
            String label = line.split("\t")[0];
            String text = line.split("\t")[1];
            float[] predicted = classifier.predict(text);
            String predicted_label = classifier.predict_label(text);
            System.out.println(text);
            System.out.println("Outcome: " + predicted[0] + ", " + predicted[1]);
            System.out.println("Predicted: " + predicted_label + " Actual: " + label);
        }
    }
}

Sentiment Analysis using Bi-directional LSTM

Below show the demo codes of the BidirectionalLstmSentimentClassifier which loads the wordvec_bidirectional_lstm.pb tensorflow model file, and uses it to do sentiment analysis:

import com.github.chen0040.tensorflow.classifiers.sentiment.models.BidirectionalLstmSentimentClassifier;
import com.github.chen0040.tensorflow.classifiers.sentiment.utils.ResourceUtils;

import java.io.IOException;
import java.util.List;

public class BidirectionalLstmSentimentClassifierDemo {
    public static void main(String[] args) throws IOException {
        BidirectionalLstmSentimentClassifier classifier = new BidirectionalLstmSentimentClassifier();
        classifier.load_model(ResourceUtils.getInputStream("tf_models/bidirectional_lstm_softmax.pb"));
        classifier.load_vocab(ResourceUtils.getInputStream("tf_models/bidirectional_lstm_softmax.csv"));

        List<String> lines = ResourceUtils.getLines("data/umich-sentiment-train.txt");

        for(String line : lines){
            String label = line.split("\t")[0];
            String text = line.split("\t")[1];
            float[] predicted = classifier.predict(text);
            String predicted_label = classifier.predict_label(text);
            System.out.println(text);
            System.out.println("Outcome: " + predicted[0] + ", " + predicted[1]);
            System.out.println("Predicted: " + predicted_label + " Actual: " + label);
        }
    }
}

Audio Classification using Cifar10 Audio Classifier

Below show the demo codes of the Cifar10AudioClassifier which loads the cifar10.pb tensorflow model file, and uses it to do music genres prediction:

import com.github.chen0040.tensorflow.classifiers.audio.models.cifar10.Cifar10AudioClassifier;
import com.github.chen0040.tensorflow.classifiers.audio.utils.ResourceUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import java.io.File;
import java.io.IOException;
import java.io.InputStream;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;

public class Cifar10AudioClassifierDemo {

    private static final Logger logger = LoggerFactory.getLogger(Cifar10AudioClassifierDemo.class);

    private static List<String> getAudioFiles() {
        List<String> result = new ArrayList<>();
        File file = new File("gtzan/genres");
        System.out.println(file.getAbsolutePath());
        if (file.isDirectory()) {
            for (File class_folder : file.listFiles()) {
                if (class_folder.isDirectory()) {
                    for (File f : class_folder.listFiles()) {
                        String file_path = f.getAbsolutePath();
                        if (file_path.endsWith("au")) {
                            result.add(file_path);

                        }
                    }
                }
            }
        }

        return result;
    }

    public static void main(String[] args) throws IOException {
        InputStream inputStream = ResourceUtils.getInputStream("tf_models/cifar10.pb");
        Cifar10AudioClassifier classifier = new Cifar10AudioClassifier();
        classifier.load_model(inputStream);

        List<String> paths = getAudioFiles();

        Collections.shuffle(paths);

        for (String path : paths) {
            System.out.println("Predicting " + path + " ...");
            File f = new File(path);
            String label = classifier.predict_audio(f);

            System.out.println("Predicted: " + label);
        }
    }
}

Audio Classification using ResNetV2 Audio Classifier

Below show the demo codes of the ResNetV2AudioClassifier which loads the resnet-v2.pb tensorflow model file, and uses it to do music genres prediction:

import com.github.chen0040.tensorflow.classifiers.audio.models.resnet.ResNetV2AudioClassifier;
import com.github.chen0040.tensorflow.classifiers.audio.utils.ResourceUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import java.io.File;
import java.io.IOException;
import java.io.InputStream;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;

public class ResNetV2AudioClassifierDemo {

    private static final Logger logger = LoggerFactory.getLogger(ResNetV2AudioClassifierDemo.class);

     private static List<String> getAudioFiles() {
        List<String> result = new ArrayList<>();
        File dir = new File("music_samples");
        System.out.println(dir.getAbsolutePath());
        if (dir.isDirectory()) {

            for (File f : dir.listFiles()) {
                String file_path = f.getAbsolutePath();
                if (file_path.endsWith("au")) {
                    result.add(file_path);

                }
            }
        }

        return result;
    }

    public static void main(String[] args) throws IOException {
        InputStream inputStream = ResourceUtils.getInputStream("tf_models/resnet-v2.pb");
        ResNetV2AudioClassifier classifier = new ResNetV2AudioClassifier();
        classifier.load_model(inputStream);

        List<String> paths = getAudioFiles();

        Collections.shuffle(paths);

        for (String path : paths) {
            System.out.println("Predicting " + path + " ...");
            File f = new File(path);
            String label = classifier.predict_audio(f);

            System.out.println("Predicted: " + label);
        }
    }
}

Audio Search Engine

The sample codes below shows how to index and search for audio file using the AudioSearchEngine class:

AudioSearchEngine searchEngine = new AudioSearchEngine();
if(!searchEngine.loadIndexDbIfExists()) {
    searchEngine.indexAll(new File("music_samples").listFiles());
    searchEngine.saveIndexDb();
}

int pageIndex = 0;
int pageSize = 20;
boolean skipPerfectMatch = true;
for(File f : new File("music_samples").listFiles()) {
    System.out.println("querying similar music to " + f.getName());
    List<AudioSearchEntry> result = searchEngine.query(f, pageIndex, pageSize, skipPerfectMatch);
    for(int i=0; i < result.size(); ++i){
        System.out.println("# " + i + ": " + result.get(i).getPath() + " (distSq: " + result.get(i).getDistanceSq() + ")");
    }
}

Music Recommend-er

The sample codes below shows how to recommend musics based on user's music history using the KnnAudioRecommender class:

AudioUserHistory userHistory = new AudioUserHistory();

List<String> audioFiles = FileUtils.getAudioFiles();
Collections.shuffle(audioFiles);

for(int i=0; i < 40; ++i){
    String filePath = audioFiles.get(i);
    userHistory.logAudio(filePath);
    try {
        Thread.sleep(100L);
    } catch (InterruptedException e) {
        e.printStackTrace();
    }
}

KnnAudioRecommender recommender = new KnnAudioRecommender();
if(!recommender.loadIndexDbIfExists()) {
    recommender.indexAll(new File("music_samples").listFiles(a -> a.getAbsolutePath().toLowerCase().endsWith(".au")));
    recommender.saveIndexDb();
}

System.out.println(userHistory.head(10));

int k = 10;
List<AudioSearchEntry> result = recommender.recommends(userHistory.getHistory(), k);

for(int i=0; i < result.size(); ++i){
    AudioSearchEntry entry = result.get(i);
    System.out.println("Search Result #" + (i+1) + ": " + entry.getPath());
}