org.canova.cli.subcommands.TestVectorize.java Source code

Java tutorial

Introduction

Here is the source code for org.canova.cli.subcommands.TestVectorize.java

Source

/*
 *
 *  *
 *  *  * Copyright 2015 Skymind,Inc.
 *  *  *
 *  *  *    Licensed under the Apache License, Version 2.0 (the "License");
 *  *  *    you may not use this file except in compliance with the License.
 *  *  *    You may obtain a copy of the License at
 *  *  *
 *  *  *        http://www.apache.org/licenses/LICENSE-2.0
 *  *  *
 *  *  *    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.
 *  *
 *
 */

package org.canova.cli.subcommands;

import static org.junit.Assert.*;

import java.io.File;
import java.io.IOException;
import java.net.URL;
import java.util.ArrayList;
import java.util.Collection;
import java.util.List;

import org.apache.commons.io.FileUtils;
import org.canova.api.conf.Configuration;
import org.canova.api.formats.input.InputFormat;
import org.canova.api.records.reader.RecordReader;
import org.canova.api.split.FileSplit;
import org.canova.api.split.InputSplit;
import org.canova.api.util.ArchiveUtils;
import org.canova.api.writable.Writable;
import org.canova.image.loader.LFWLoader;
import org.canova.image.recordreader.MNISTRecordReader;
import org.junit.Test;

public class TestVectorize {

    private static final String trainingFilesFilename = "images-idx1-ubyte.gz";
    private static final String trainingFilesURL = "http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz";
    public static final String trainingFilesFilename_unzipped = "images-idx1-ubyte";

    private static final String trainingFileLabelsURL = "http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz";
    private static final String trainingFileLabelsFilename = "labels-idx1-ubyte.gz";
    public static final String trainingFileLabelsFilename_unzipped = "labels-idx1-ubyte";

    static String TEMP_ROOT = "/tmp"; //System.getProperty("user.home");
    static String MNIST_ROOT = TEMP_ROOT + File.separator + "MNIST" + File.separator;

    static String MNIST_Filename = MNIST_ROOT + MNISTRecordReader.trainingFilesFilename_unzipped;

    public void checkForMNISTLocally() {

        // 1. check for the MNIST data first!

        // does it exist?

        // if not, then let's download it

        System.out.println("Checking to see if MNIST exists locally: " + MNIST_ROOT);

        if (!new File(MNIST_ROOT).exists()) {

            new File(MNIST_ROOT).mkdir();

            System.out.println("Downloading and unzipping the MNIST dataset locally to: " + MNIST_ROOT);
            try {
                //new MnistFetcher().downloadAndUntar();
                downloadAndUntar();
            } catch (IOException e) {
                // TODO Auto-generated catch block
                e.printStackTrace();
            }
        } else {

            System.out.println("MNIST already exists locally...");

        }

        if (new File(MNIST_Filename).exists()) {
            System.out.println("The images file exists locally unzipped!");
        } else {
            System.out.println("The images file DOES NOT exist locally unzipped!");
            System.out.println("Downloading and unzipping the MNIST dataset locally to: " + MNIST_ROOT);
            try {
                //new MnistFetcher().downloadAndUntar();
                downloadAndUntar();
            } catch (IOException e) {
                // TODO Auto-generated catch block
                e.printStackTrace();
            }

        }

    }

    /**
     * Added another copy of this method because MNISTFetcher uses the user's system home directory
     * and we couldnt specify that in the canova conf file
     *
     * @return
     * @throws IOException
     */
    public static File downloadAndUntar() throws IOException {

        File fileDir;

        //      if(fileDir != null) {
        //      return fileDir;
        //}
        // mac gives unique tmp each run and we want to store this persist
        // this data across restarts
        //File tmpDir = new File(System.getProperty("user.home"));

        File baseDir = new File(MNIST_ROOT);
        if (!(baseDir.isDirectory() || baseDir.mkdir())) {
            throw new IOException("Could not mkdir " + baseDir);
        }

        //log.info("Downloading mnist...");
        // getFromOrigin training records
        File tarFile = new File(baseDir, trainingFilesFilename);

        if (!tarFile.isFile()) {
            FileUtils.copyURLToFile(new URL(trainingFilesURL), tarFile);
        }

        ArchiveUtils.unzipFileTo(tarFile.getAbsolutePath(), baseDir.getAbsolutePath());

        // getFromOrigin training records
        File labels = new File(baseDir, trainingFileLabelsFilename);

        if (!labels.isFile()) {
            FileUtils.copyURLToFile(new URL(trainingFileLabelsURL), labels);
        }

        ArchiveUtils.unzipFileTo(labels.getAbsolutePath(), baseDir.getAbsolutePath());

        fileDir = baseDir;
        return fileDir;
    }

    public static File download_LFW_AndUntar(String workingBaseDir) throws Exception {
        new LFWLoader().load();
        FileUtils.copyDirectory(new File(System.getProperty("user.home"), "lfw"),
                new File(System.getProperty("java.io.tmpdir"), "lfw"));
        return new File(System.getProperty("java.io.tmpdir"), "lfw");
    }

    /**
     * Creates an input format
     *
     * @return
     */
    public static InputFormat createInputFormat(String inputFormat) {
        try {
            Class<? extends InputFormat> inputFormatClazz = (Class<? extends InputFormat>) Class
                    .forName(inputFormat);
            return inputFormatClazz.newInstance();
        } catch (Exception e) {
            throw new RuntimeException(e);
        }

    }

    public static int checkNumberOfRecordsInSVMLightOutput(String filenamePath)
            throws IOException, InterruptedException {

        //String[] args = { "-conf", "src/test/resources/text/conf/text_vectorization_conf_unit_test.txt" };
        //Vectorize vecCommand = new Vectorize( args );

        Configuration conf = new Configuration();
        //        conf.set( OutputFormat.OUTPUT_PATH, "" );

        String inputFormatKey = "input.format";
        String svmLightInputFormat = "org.canova.api.formats.input.impl.SVMLightInputFormat";

        conf.set(inputFormatKey, svmLightInputFormat);

        //String datasetInputPath = (String) vecCommand.configProps.get("input.directory");

        InputFormat inputformat = createInputFormat(svmLightInputFormat);

        //RecordReader rr = inputformat.

        File inputFile = new File(filenamePath);
        InputSplit split = new FileSplit(inputFile);
        //InputFormat inputFormat = this.createInputFormat();

        System.out.println("input file: " + filenamePath);

        RecordReader reader = inputformat.createReader(split, conf);

        int count = 0;
        while (reader.hasNext()) {

            count++;
            Collection<Writable> vector = reader.next();

            String label = getLabelFromSVMLightVector(vector);
            //System.out.println( label );

        }

        //assertEquals( 4, count );

        return count;

    }

    public static int countLabelsInSVMLightOutput(String filenamePath) throws IOException, InterruptedException {

        List<String> labels = new ArrayList<>();

        Configuration conf = new Configuration();

        String inputFormatKey = "input.format";
        String svmLightInputFormat = "org.canova.api.formats.input.impl.SVMLightInputFormat";

        conf.set(inputFormatKey, svmLightInputFormat);

        //String datasetInputPath = (String) vecCommand.configProps.get("input.directory");

        InputFormat inputformat = createInputFormat(svmLightInputFormat);

        //RecordReader rr = inputformat.

        File inputFile = new File(filenamePath);
        InputSplit split = new FileSplit(inputFile);
        //InputFormat inputFormat = this.createInputFormat();

        System.out.println("input file: " + filenamePath);

        RecordReader reader = inputformat.createReader(split, conf);

        //int count = 0;
        while (reader.hasNext()) {

            //count++;
            Collection<Writable> vector = reader.next();

            String labelName = getLabelFromSVMLightVector(vector);
            //System.out.println( label );

            if (!labels.contains(labelName)) {
                labels.add(labelName);
            }

        }

        reader.close();
        return labels.size();
        //assertEquals( 4, count );

        //return count;

    }

    public static String getLabelFromSVMLightVector(Collection<Writable> vector) {
        return vector.toArray()[vector.size() - 1].toString();
    }

    public static void setupLFWSampleLocally() throws Exception {

        String localUnzippedSubdir = "lfw";
        String workingDir = "/tmp/canova/image/"; // + localUnzippedSubdir;

        // does the file exist locally?

        download_LFW_AndUntar(workingDir);

        // let's only get a few images in 2 labels

    }

    @Test
    public void testLoadConfFile() throws IOException {

        String[] args = { "-conf", "src/test/resources/csv/confs/unit_test_conf.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.loadConfigFile();
        assertEquals("/tmp/iris_unit_test_sample.txt", vecCommand.configProps.get("canova.output.directory"));

    }

    @Test
    public void testExecuteCSVConversionWorkflow() throws Exception {

        String[] args = { "-conf", "src/test/resources/csv/confs/unit_test_conf.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.execute();

        // now check the output
        int count = checkNumberOfRecordsInSVMLightOutput(vecCommand.outputVectorFilename);
        assertEquals(12, count);
    }

    @Test
    public void testExecuteCSVConversionWorkflow_WithShuffle() throws Exception {

        String[] args = { "-conf", "src/test/resources/csv/confs/unit_test_csv_conf_w_shuffle.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.execute();

        // now check the output
        int count = checkNumberOfRecordsInSVMLightOutput(vecCommand.outputVectorFilename);
        assertEquals(12, count);
    }

    @Test
    public void testExecuteCSVConversionWorkflow_SkipHeader() throws Exception {

        String[] args = { "-conf", "src/test/resources/csv/confs/unit_test_csv_conf_skip_header.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.execute();

        // now check the output
        int count = checkNumberOfRecordsInSVMLightOutput(vecCommand.outputVectorFilename);
        assertEquals(12, count);
    }

    @Test
    public void testExecuteImageCustomMNISTInputFormatConversionWorkflow() throws Exception {

        checkForMNISTLocally();

        String[] args = { "-conf", "src/test/resources/image/conf/mnist/unit_test_conf.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.execute();

        // now check the output

        // 1. how many labels are there?

        // 2. how many vectors are in the output?

        int count = checkNumberOfRecordsInSVMLightOutput(vecCommand.outputVectorFilename);
        int labelCount = countLabelsInSVMLightOutput(vecCommand.outputVectorFilename);

        //System.out.println( "Vectors in file: " + count );

        // this seems .... odd?
        assertEquals(59999, count);
        assertEquals(10, labelCount);

    }

    /**
     * Testing the normal image input format reader
     *
     * should be 530 records in the output
     *
     *
     * @throws Exception
     */
    @Test
    public void testExecuteImageInputFormatConversionWorkflow() throws Exception {

        //   1. setup the LWF dataset sample to work with

        setupLFWSampleLocally();

        String[] args = { "-conf", "src/test/resources/image/conf/unit_test_conf.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.execute();

        // now check the output

        // 1. how many labels are there?

        // 2. how many vectors are in the output?

        int count = checkNumberOfRecordsInSVMLightOutput(vecCommand.outputVectorFilename);
        int labelCount = countLabelsInSVMLightOutput(vecCommand.outputVectorFilename);

        System.out.println("Vectors in file: " + count);

        assertEquals(13233, count);
        assertEquals(5749, labelCount);

    }

    /**
     * Testing the normal image input format reader
     *
     * should be 530 records in the output
     *
     * with shuffle!
     *
     * @throws Exception
     */
    @Test
    public void testExecuteImageInputFormat_WithShuffle_ConversionWorkflow() throws Exception {

        //   1. setup the LWF dataset sample to work with

        setupLFWSampleLocally();

        String[] args = { "-conf", "src/test/resources/image/conf/unit_test_w_shuffle_conf.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.execute();

        // now check the output

        // 1. how many labels are there?

        // 2. how many vectors are in the output?

        int count = checkNumberOfRecordsInSVMLightOutput(vecCommand.outputVectorFilename);
        int labelCount = countLabelsInSVMLightOutput(vecCommand.outputVectorFilename);

        System.out.println("Vectors in file: " + count);
        assertEquals(13233, count);
        assertEquals(5749, labelCount);

    }

    /**
     *
     *
     *
     * @throws Exception
     */
    @Test
    public void testExecuteTextInputFormat_TFIDF_ConversionWorkflow() throws Exception {

        String[] args = { "-conf",
                "src/test/resources/text/DemoTextFiles/conf/text_vectorization_conf_unit_test.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.execute();

        // now check the output

        // now check the output

        // 1. how many labels are there?

        // 2. how many vectors are in the output?

        int count = checkNumberOfRecordsInSVMLightOutput(vecCommand.outputVectorFilename);
        int labelCount = countLabelsInSVMLightOutput(vecCommand.outputVectorFilename);

        //   System.out.println( "Vectors in file: " + count );

        // this seems .... odd?
        assertEquals(4, count);
        assertEquals(2, labelCount);

        // check the vocab?

    }

    @Test
    public void testExecuteTextInputFormat_TFIDF_Tweets_ConversionWorkflow() throws Exception {

        String[] args = { "-conf", "src/test/resources/text/Tweets/conf/tweet_conf.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.execute();

        // 1. how many labels are there?

        // 2. how many vectors are in the output?

        int count = checkNumberOfRecordsInSVMLightOutput(vecCommand.outputVectorFilename);
        int labelCount = countLabelsInSVMLightOutput(vecCommand.outputVectorFilename);

        //   System.out.println( "Vectors in file: " + count );

        // this seems .... odd?
        assertEquals(15, count);
        assertEquals(3, labelCount);

        // check the vocab?

    }

    @Test
    public void testExecuteTextInputFormat_TFIDF_Tweets_ConversionWorkflow_WithShuffle() throws Exception {

        String[] args = { "-conf", "src/test/resources/text/Tweets/conf/tweet_conf_w_shuffle.txt" };
        Vectorize vecCommand = new Vectorize(args);

        vecCommand.execute();

        // 1. how many labels are there?

        // 2. how many vectors are in the output?

        int count = checkNumberOfRecordsInSVMLightOutput(vecCommand.outputVectorFilename);
        int labelCount = countLabelsInSVMLightOutput(vecCommand.outputVectorFilename);

        //   System.out.println( "Vectors in file: " + count );

        // this seems .... odd?
        assertEquals(15, count);
        assertEquals(3, labelCount);

        // check the vocab?

    }

    /**
     * TODO: need some work here on the LFW test data
     *
     *
     * @throws Exception
     */
    @Test
    public void testExecuteAudioInputFormatConversionWorkflow() throws Exception {

        //   1. setup the LWF dataset sample to work with

        //   setupLFWSampleLocally();
        /*
        String[] args = { "-conf", "src/test/resources/audio/conf/unit_test_conf.txt" };      
        Vectorize vecCommand = new Vectorize( args );
            
        vecCommand.execute();
        */
        // now check the output

        // 1. how many labels are there?

        // 2. how many vectors are in the output?

    }

}