/**
 * Copyright (c) 2011, The University of Southampton and the individual contributors.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without modification,
 * are permitted provided that the following conditions are met:
 *
 *   *  Redistributions of source code must retain the above copyright notice,
 *      this list of conditions and the following disclaimer.
 *
 *   *  Redistributions in binary form must reproduce the above copyright notice,
 *      this list of conditions and the following disclaimer in the documentation
 *      and/or other materials provided with the distribution.
 *
 *   *  Neither the name of the University of Southampton nor the names of its
 *      contributors may be used to endorse or promote products derived from this
 *      software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
 * ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */
package org.openimaj.classifier.citylandscape;

import java.io.BufferedReader;
import java.io.File;
import java.io.IOException;
import java.io.InputStream;
import java.io.InputStreamReader;

import org.openimaj.image.FImage;
import org.openimaj.image.ImageUtilities;
import org.openimaj.image.analysis.algorithm.EdgeDirectionCoherenceVector;
import org.openimaj.knn.DoubleNearestNeighbours;
import org.openimaj.knn.approximate.DoubleNearestNeighboursKDTree;

/**
 *      Originally written by Ajay Mehta for his third year project. Reworked for
 *      inclusion into OpenIMAJ by David Dupplaw.
 *      <p>
 *      This class provides classification of objects. The idea in the classifier
 *      is to classfy images between landscape images and cityscape images. Could
 *      also be used as a natural vs. non-natural classifier as the technique used
 *      is based on edge-direction coherence vectors (i.e. looking for strong lines).
 *      The training set (src/main/resources/CityLS10000.2.no-decimal) 
 *      provides 5000 examples of edge direction coherence vectors for both city
 *      scape and landscape (the original images were crawled from Flickr based on
 *      appropriate tags). Internally the classifier uses the nearest neighbour
 *      to determine the class of the query.
 *      <p>
 *      To use, do something like this:
 *      <pre>
 *      {@code
 *              CityLandscapeTree clt = new CityLandscapeTree( "city", "landscape",
 *                      getClass().getResourceAsStream( "/CityLS10000.2.no-decimal" ), 10000 );
 *              String clazz = clt.classifyImage( ImageUtilities.readF( 
 *                      new File("myImage.jpg") ), 1 ); 
 *      }
 *      </pre>
 *      <p>
 *      A main method is supplied for the tool which will take an image filename
 *      and classify the image as city or landscape.
 * 
 *      @author Ajay Mehta (am24g08@ecs.soton.ac.uk)
 *      @author David Dupplaw (dpd@ecs.soton.ac.uk)
 *      @created 2011
 *      
 */
public class CityLandscapeTree
{
        /** Category 1 */
        private String cat1 = null;

        /** Category 2 */
        private String cat2 = null;

        /** The number of instances in the training set */
        private int trainingSetSize = 0;

        /** The nearest neighbour classifier */
        private DoubleNearestNeighbours dnn = null;
        
        /** The size of the vector generated for each image */
        private static int VECTOR_SIZE = 144;

        /** */
        private final static int NTREES = 768;

        /** */
        private final static int NCHECKS = 8;

        /**
         *      Default constructor that takes the two categories
         *      to classifier between.
         * 
         *      @param cat1 The first category
         *      @param cat2 The second category
         */
        public CityLandscapeTree( String cat1, String cat2 )
        {
                this.cat1 = cat1;
                this.cat2 = cat2;
        }
        
        /**
         *      Constructor that takes the two categories to classify between,
         *      an input stream that points to a training set file and the size
         *      of that training set. 
         * 
         *      @param cat1 The first classification category
         *      @param cat2 The second classification category
         *      @param trainingSet The training set
         *      @param trainingSetSize The size of the training set
         */
        public CityLandscapeTree( String cat1, String cat2,
                        InputStream trainingSet, int trainingSetSize )
        {
                this( cat1, cat2 );
                try
                {
                        // Load the given training set
                        double[][] train = loadTrainingSet( trainingSet, trainingSetSize );
                        
                        // Construct DoubleNearestNeighbours Object
                        dnn = new DoubleNearestNeighboursKDTree( train, NTREES, NCHECKS );
                }
                catch( IOException e )
                {
                        e.printStackTrace();
                }
        }

        /**
         *      From the given image filename, returns the edge direction coherence
         *      vector as a double array.
         * 
         *      @param imageName The image to process
         *      @return A 2-dimensional vector with only one 
         */
        private static double[][] getImageVectorAsArray( FImage crgbimage )
        {
                double[][] toReturn = new double[1][VECTOR_SIZE];
                
                // Calculate the Edge direction coherence on the image.
                EdgeDirectionCoherenceVector edcv = new EdgeDirectionCoherenceVector();
                edcv.setNumberOfBins( VECTOR_SIZE/2 );
                
                // Process the image
                crgbimage.analyseWith( edcv );
                
                // Get the histogram
                double[] d = edcv.getLastHistogram().asDoubleFV().asDoubleVector();
                
                // Normalise the vector by the total number of edge pixels
                double[] edgeCounter = new double[1];
                for( int j = 0; j < VECTOR_SIZE; j++ )
                {
                        toReturn[0][j]  = d[j];
                        edgeCounter[0] += d[j];
                }
                
                // Normalise the vector
                CityLandscapeUtilities.normaliseVector( toReturn, edgeCounter );

                return toReturn;
        }

        /**
         *      Classifies the given image.
         * 
         *      @param image The image to classify
         *      @param k The number of nearest neighbours to interrogate
         *      @return The expected category
         */
        public String classifyImage( FImage image, int k )
        {
                System.out.println( "Classifying image... " );
                
                // Get the vector for the query image.
                double[][] query = getImageVectorAsArray( image );
                
                // Indexes and distances of nearest neighbours for the one query image
                int[][] indexes = new int[1][k];
                double[][] distances = new double[1][k];
                
                // KNN search for the query image
                dnn.searchKNN( query, k, indexes, distances );
                
                // Counters for the two categories (City vs Landscape)
                double cat1Counter = 0, cat2Counter = 0;

                // Loop through all the results
                for( int i = 0; i < distances[0].length; i++ )
                {
                        // 
                        if( indexes[0][i] < trainingSetSize / 2 )
                                        cat1Counter += 1 / distances[0][i];
                        else    cat2Counter += 1 / distances[0][i];
                }

                if( cat1Counter > cat2Counter )
                                return cat1;
                else if( cat2Counter > cat1Counter )
                                return cat2;
                else    return "?";
        }

        /**
         *      Loads a training set of a given size from an input stream
         * 
         *      @param is The input stream to read the training set from
         *      @param vecSize The size of the training set
         *      @return A 2-dimensional double array of the training set
         *      @throws IOException if the input stream could not be fully read
         */
        public double[][] loadTrainingSet( InputStream is, int vecSize ) 
                throws IOException
        {
                this.trainingSetSize = vecSize;
                double[][] trainingVector = new double[trainingSetSize][VECTOR_SIZE];
                double[] totalEdges = new double[trainingSetSize];
                
                System.out.println( "Loading training data... " );
                
                // Read in each line from the training set data. The training set data
                // is set out as one-line per training image. The data in the line is
                // comma-separated double values, where the values are the histogram
                // bin values from the edge direction coherence vector. The number
                // of bins should be this.vectorSize/2 each for coherence and incoherent
                // histograms. A final value on each line contains the total number
                // of detected coherent edges in the image.
                BufferedReader br = new BufferedReader( new InputStreamReader( is ) );          
                int counter = 0;
                String line = null;
                while( (line = br.readLine()) != null )
                {
                        String[] array = line.split( "," );

                        for( int i = 0; i < VECTOR_SIZE/2; i++ )
                                trainingVector[counter][i] = Double.parseDouble( array[i] );
                        
                        totalEdges[counter++] = Double.parseDouble( 
                                        array[array.length - 1] );
                }

                // Normalise the vector by the total number of edges.
                System.out.println( "Normalising training data..." );
                CityLandscapeUtilities.normaliseVector( trainingVector, totalEdges );
                
                return trainingVector;
        }
        
        /**
         *      Given an image filename, it will classify it.
         *  @param args
         */
        public static void main( String[] args )
    {
            if( args.length < 1 )
            {
                System.err.println( "Please supply an image filename." );
                System.exit(1);
            }
            
            try
        {
                CityLandscapeTree clt = new CityLandscapeTree( "City", "Landscape",
                                CityLandscapeTree.class.getResourceAsStream( "/CityLS10000.2.no-decimal" ), 10000 );
                String clazz = clt.classifyImage( ImageUtilities.readF( new File(args[0]) ), 1 );
                System.out.println( "Classified as: "+clazz );
        }
        catch( IOException e )
        {
                e.printStackTrace();
        }
    }
}