/*
* This file is part of the LIRE project: http://www.semanticmetadata.net/lire
* LIRE is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* LIRE is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with LIRE; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* We kindly ask you to refer the any or one of the following publications in
* any publication mentioning or employing Lire:
*
* Lux Mathias, Savvas A. Chatzichristofis. Lire: Lucene Image Retrieval –
* An Extensible Java CBIR Library. In proceedings of the 16th ACM International
* Conference on Multimedia, pp. 1085-1088, Vancouver, Canada, 2008
* URL: http://doi.acm.org/10.1145/1459359.1459577
*
* Lux Mathias. Content Based Image Retrieval with LIRE. In proceedings of the
* 19th ACM International Conference on Multimedia, pp. 735-738, Scottsdale,
* Arizona, USA, 2011
* URL: http://dl.acm.org/citation.cfm?id=2072432
*
* Mathias Lux, Oge Marques. Visual Information Retrieval using Java and LIRE
* Morgan & Claypool, 2013
* URL: http://www.morganclaypool.com/doi/abs/10.2200/S00468ED1V01Y201301ICR025
*
* Copyright statement:
* ====================
* (c) 2002-2013 by Mathias Lux (mathias@juggle.at)
* http://www.semanticmetadata.net/lire, http://www.lire-project.net
*
* Updated: 27.06.14 13:06
*/
package net.semanticmetadata.lire.benchmarking;
import junit.framework.TestCase;
import net.semanticmetadata.lire.DocumentBuilder;
import net.semanticmetadata.lire.ImageSearchHits;
import net.semanticmetadata.lire.ImageSearcher;
import net.semanticmetadata.lire.imageanalysis.bovw.BOVWBuilder;
import net.semanticmetadata.lire.imageanalysis.bovw.SimpleFeatureBOVWBuilder;
import net.semanticmetadata.lire.imageanalysis.sift.Feature;
import net.semanticmetadata.lire.imageanalysis.spatialpyramid.SPACC;
import net.semanticmetadata.lire.imageanalysis.spatialpyramid.SPCEDD;
import net.semanticmetadata.lire.imageanalysis.spatialpyramid.SPFCTH;
import net.semanticmetadata.lire.imageanalysis.spatialpyramid.SPJCD;
import net.semanticmetadata.lire.impl.*;
import net.semanticmetadata.lire.indexing.parallel.ParallelIndexer;
import net.semanticmetadata.lire.utils.FileUtils;
import org.apache.lucene.document.Document;
import org.apache.lucene.index.DirectoryReader;
import org.apache.lucene.index.IndexReader;
import org.apache.lucene.index.MultiFields;
import org.apache.lucene.store.FSDirectory;
import org.apache.lucene.store.IOContext;
import org.apache.lucene.store.RAMDirectory;
import org.apache.lucene.util.Bits;
import javax.imageio.ImageIO;
import java.io.*;
import java.util.*;
/**
* Created with IntelliJ IDEA.
* User: mlux
* Date: 14.05.13
* Time: 10:56
*/
public class UCIDBenchmark extends TestCase {
private String indexPath = "ucid-index";
// if you don't have the images you can get them here: http://homepages.lboro.ac.uk/~cogs/datasets/ucid/ucid.html
// I converted all images to PNG (lossless) to save time, space & troubles with Java.
private String testExtensive = "testdata/UCID";
private final String groundTruth = "testdata/ucid.v2.groundtruth.txt";
// private String testExtensive = "testdata/UCID.small";
// private final String groundTruth = "testdata/ucid.v2.groundtruth.small.txt";
//
private ChainedDocumentBuilder builder;
private HashMap<String, List<String>> queries;
private HashMap<String, Integer> query2id;
ParallelIndexer parallelIndexer;
/* configure test of global features by adding them here */
LireFeature[] globalFeaturesToTest = new LireFeature[]{
// new CEDD(),
// new PHOG(),
// new SPACC(),
// new SPCEDD(),
// new OpponentHistogram()
};
LireFeature[] simpleFeaturesToTest = new LireFeature[]{
new CEDD()
// new ScalableColor()
// new OpponentHistogram()
};
boolean testSift = false;
boolean testSurf = false;
protected void setUp() throws Exception {
super.setUp();
// indexPath = "ucid-index-573374558";
indexPath += "-" + System.currentTimeMillis() % (1000 * 60 * 60 * 24 * 7);
// Setting up DocumentBuilder:
parallelIndexer = new ParallelIndexer(16, indexPath, testExtensive, true) {
@Override
public void addBuilders(ChainedDocumentBuilder builder) {
for (int i = 0; i < globalFeaturesToTest.length; i++) {
LireFeature lireFeature = globalFeaturesToTest[i];
builder.addBuilder(new GenericDocumentBuilder(lireFeature.getClass()));
}
for (int i = 0; i < simpleFeaturesToTest.length; i++) {
LireFeature lireFeature = simpleFeaturesToTest[i];
// need to create a feature instance for each thread.
try {
builder.addBuilder(new SimpleBuilder(lireFeature.getClass().newInstance()));
} catch (Exception e) {
e.printStackTrace();
}
}
if (testSift) builder.addBuilder(new SiftDocumentBuilder());
if (testSurf) builder.addBuilder(new SurfDocumentBuilder());
}
};
// Getting the queries:
BufferedReader br = new BufferedReader(new FileReader(groundTruth));
String line;
queries = new HashMap<String, List<String>>(260);
query2id = new HashMap<String, Integer>(260);
int qID = 1;
String currentQuery = null;
LinkedList<String> results = null;
while ((line = br.readLine()) != null) {
line = line.trim();
if (line.startsWith("#") || line.length() < 4)
continue;
else {
if (line.endsWith(":")) {
if (currentQuery != null) {
queries.put(currentQuery, results);
query2id.put(currentQuery, qID);
qID++;
}
currentQuery = line.replace(':', ' ').trim();
results = new LinkedList<String>();
} else {
results.add(line);
}
}
}
queries.put(currentQuery, results);
query2id.put(currentQuery, qID);
}
public void testMAP() throws IOException {
// ************************* INDEXING *************************
parallelIndexer.run();
if (testSurf) {
System.out.println("** SURF BoVW");
BOVWBuilder sh = new BOVWBuilder(DirectoryReader.open(FSDirectory.open(new File(indexPath))), new SurfFeature(), 500, 128);
sh.index();
}
if (testSift) {
System.out.println("** SIFT BoVW");
BOVWBuilder sh = new BOVWBuilder(DirectoryReader.open(FSDirectory.open(new File(indexPath))), new Feature(), 500, 128);
sh.index();
}
for (int i = 0; i < simpleFeaturesToTest.length; i++) {
LireFeature lireFeature = simpleFeaturesToTest[i];
System.out.println("** SIMPLE BoVW with " + lireFeature.getFeatureName());
SimpleFeatureBOVWBuilder ldb = new SimpleFeatureBOVWBuilder(DirectoryReader.open(FSDirectory.open(new File(indexPath))), lireFeature, SimpleBuilder.KeypointDetector.CVSURF, 500, 128);
ldb.index();
}
// System.out.println("** SIMPLE BoVW / LoDe CEDD");
// ldb = new SimpleFeatureBOVWBuilder(DirectoryReader.open(FSDirectory.open(new File(indexPath))), 500, 128, new CEDD());
// ldb.index();
// VLAD
// VLADBuilder vladBuilder = new VLADBuilder(DirectoryReader.open(FSDirectory.open(new File(indexPath))));
// vladBuilder.index();
// ************************* SEARCHING *************************
IndexReader reader = DirectoryReader.open(new RAMDirectory(FSDirectory.open(new File(indexPath)), IOContext.READONCE));
System.out.println("Feature\tMAP\tp@10\tER");
for (int i = 0; i < globalFeaturesToTest.length; i++) {
LireFeature lireFeature = globalFeaturesToTest[i];
computeMAP(new GenericFastImageSearcher(1400, lireFeature.getClass(), true, reader), lireFeature.getFeatureName(), reader);
}
// computeMAP(new VisualWordsImageSearcher(1400, DocumentBuilder.FIELD_NAME_SURF + DocumentBuilder.FIELD_NAME_BOVW), "Surf BoVW Lucene", reader);
if (testSurf)
// computeMAP(new GenericFastImageSearcher(1400, GenericDoubleLireFeature.class, DocumentBuilder.FIELD_NAME_SURF_LOCAL_FEATURE_HISTOGRAM, true, reader), "Surf BoVW L2", reader);
computeMAP(new GenericFastImageSearcher(1400, GenericDoubleLireFeature.class, DocumentBuilder.FIELD_NAME_SURF + DocumentBuilder.FIELD_NAME_BOVW_VECTOR, true, reader), "Surf BoVW L2", reader);
if (testSift)
// computeMAP(new GenericFastImageSearcher(1400, GenericDoubleLireFeature.class, DocumentBuilder.FIELD_NAME_SIFT_LOCAL_FEATURE_HISTOGRAM, true, reader), "Sift BoVW L2", reader);
computeMAP(new GenericFastImageSearcher(1400, GenericDoubleLireFeature.class, DocumentBuilder.FIELD_NAME_SIFT + DocumentBuilder.FIELD_NAME_BOVW_VECTOR, true, reader), "Sift BoVW L2", reader);
for (int i = 0; i < simpleFeaturesToTest.length; i++) {
LireFeature lireFeature = simpleFeaturesToTest[i];
computeMAP(new GenericFastImageSearcher(1400, GenericDoubleLireFeature.class, lireFeature.getFieldName() + "LoDe_Hist", true, reader), "LoDe L2 " + lireFeature.getFeatureName(), reader);
}
// computeMAP(new VisualWordsImageSearcher(1400, (new ScalableColor()).getFieldName() + "LoDe"), "LoDe SC Lucene", reader);
// computeMAP(new GenericFastImageSearcher(1400, GenericDoubleLireFeature.class, (new ScalableColor()).getFieldName() + "LoDe_Hist", true, reader), "LoDe SC L2", reader);
// computeMAP(new VisualWordsImageSearcher(1400, (new CEDD()).getFieldName() + "LoDe"), "LoDe CEDD Lucene", reader);
// computeMAP(new GenericFastImageSearcher(1400, GenericDoubleLireFeature.class, (new CEDD()).getFieldName() + "LoDe_Hist", true, reader), "LoDe CEDD L2", reader);
// computeMAP(new GenericFastImageSearcher(1400, GenericByteLireFeature.class, DocumentBuilder.FIELD_NAME_SURF_VLAD, true, reader), "VLAD-SURF", reader);
}
private void computeMAP(ImageSearcher searcher, String prefix, IndexReader reader) throws IOException {
double queryCount = 0d;
double errorRate = 0;
double map = 0;
double p10 = 0;
// Needed for check whether the document is deleted.
Bits liveDocs = MultiFields.getLiveDocs(reader);
PrintWriter fw = new PrintWriter(new File("eval/" + prefix.replace(' ', '_') + "-eval.txt"));
Hashtable<Integer, String> evalText = new Hashtable<Integer, String>(260);
for (int i = 0; i < reader.maxDoc(); i++) {
if (reader.hasDeletions() && !liveDocs.get(i)) continue; // if it is deleted, just ignore it.
String fileName = getIDfromFileName(reader.document(i).getValues(DocumentBuilder.FIELD_NAME_IDENTIFIER)[0]);
if (queries.keySet().contains(fileName)) {
String tmpEval = "";
queryCount += 1d;
// ok, we've got a query here for a document ...
Document queryDoc = reader.document(i);
ImageSearchHits hits = searcher.search(queryDoc, reader);
double rank = 0;
double avgPrecision = 0;
double found = 0;
double tmpP10 = 0;
Locale.setDefault(Locale.US);
for (int y = 0; y < hits.length(); y++) {
String hitFile = getIDfromFileName(hits.doc(y).getValues(DocumentBuilder.FIELD_NAME_IDENTIFIER)[0]);
// TODO: Sort by query ID!
tmpEval += String.format(Locale.US, "%d 1 %s %d %.2f test\n", query2id.get(fileName), hitFile.substring(0, hitFile.lastIndexOf('.')), (int) rank + 1, 100 - hits.score(y));
// if (!hitFile.equals(fileName)) {
rank++;
if (queries.get(fileName).contains(hitFile) || hitFile.equals(fileName)) { // it's a hit.
found++;
// TODO: Compute error rate, etc. here.
avgPrecision += found / rank;// * (1d/queries.get(fileName).size());
// if (rank<=60) System.out.print('X');
if (rank <= 10) tmpP10++;
} else { // nothing has been found.
if (rank == 1) errorRate += 1d;
// if (rank<=60) System.out.print('-');
}
}
// }
// System.out.println();
if (found - queries.get(fileName).size() == 1)
avgPrecision /= (double) (1d + queries.get(fileName).size());
else {
// some of the results have not been found. We have to deal with it ...
System.err.println("Did not find result ;(");
}
// assertTrue(found - queries.get(fileName).size() == 0);
map += avgPrecision;
p10 += tmpP10;
evalText.put(query2id.get(fileName), tmpEval);
}
}
for (int i = 0; i < query2id.size(); i++) {
fw.write(evalText.get(i + 1));
}
fw.close();
errorRate = errorRate / queryCount;
map = map / queryCount;
p10 = p10 / (queryCount * 10d);
System.out.print(prefix);
System.out.format("\t%.5f\t%.5f\t%.5f\n", map, p10, errorRate);
}
private String getIDfromFileName(String path) {
// That's the one for Windows. Change for Linux ...
return path.substring(path.lastIndexOf('\\') + 1).replace(".jpg", ".tif");
}
public void testIndexingSpeed() throws IOException {
ArrayList<String> images = FileUtils.getAllImages(new File(testExtensive), false);
// testFeatureSpeed(images, new AutoColorCorrelogram());
// testFeatureSpeed(images, new CEDD());
// testFeatureSpeed(images, new FCTH());
// testFeatureSpeed(images, new JCD());
testFeatureSpeed(images, new SPACC());
testFeatureSpeed(images, new SPCEDD());
testFeatureSpeed(images, new SPFCTH());
testFeatureSpeed(images, new SPJCD());
}
public void testSearchSpeed() throws IOException {
ArrayList<String> images = FileUtils.getAllImages(new File(testExtensive), false);
testSearchSpeed(images, AutoColorCorrelogram.class);
testSearchSpeed(images, CEDD.class);
testSearchSpeed(images, FCTH.class);
testSearchSpeed(images, JCD.class);
testSearchSpeed(images, SPACC.class);
testSearchSpeed(images, SPCEDD.class);
testSearchSpeed(images, SPFCTH.class);
testSearchSpeed(images, SPJCD.class);
}
private void testSearchSpeed(ArrayList<String> images, final Class featureClass) throws IOException {
parallelIndexer = new ParallelIndexer(8, indexPath, testExtensive, true) {
@Override
public void addBuilders(ChainedDocumentBuilder builder) {
builder.addBuilder(new GenericDocumentBuilder(featureClass, "feature"));
}
};
parallelIndexer.run();
IndexReader reader = DirectoryReader.open(new RAMDirectory(FSDirectory.open(new File(indexPath)), IOContext.READONCE));
Bits liveDocs = MultiFields.getLiveDocs(reader);
double queryCount = 0d;
ImageSearcher searcher = new GenericFastImageSearcher(100, featureClass, "feature");
long ms = System.currentTimeMillis();
for (int i = 0; i < reader.maxDoc(); i++) {
if (reader.hasDeletions() && !liveDocs.get(i)) continue; // if it is deleted, just ignore it.
String fileName = getIDfromFileName(reader.document(i).getValues(DocumentBuilder.FIELD_NAME_IDENTIFIER)[0]);
if (queries.keySet().contains(fileName)) {
queryCount += 1d;
// ok, we've got a query here for a document ...
Document queryDoc = reader.document(i);
ImageSearchHits hits = searcher.search(queryDoc, reader);
}
}
ms = System.currentTimeMillis() - ms;
System.out.printf("%s \t %3.1f \n", featureClass.getName().substring(featureClass.getName().lastIndexOf('.') + 1), (double) ms / queryCount);
}
private void testFeatureSpeed(ArrayList<String> images, LireFeature feature) throws IOException {
long ms = System.currentTimeMillis();
for (Iterator<String> iterator = images.iterator(); iterator.hasNext(); ) {
String s = iterator.next();
feature.extract(ImageIO.read(new File(s)));
}
ms = System.currentTimeMillis() - ms;
System.out.printf("%s \t %3.1f \n", feature.getClass().getName().substring(feature.getClass().getName().lastIndexOf('.') + 1), (double) ms / (double) images.size());
}
}