def main():
import sys
sys.path.append('../Utils')
sys.path.append('../Utils/Evaluation')
sys.path.append('../Utils/Tuning')
import Matrix as matrix
import Load as load
import numpy,learningCurves
import measures
import SVM_Tuning
trainPath = '/home/leonidas/Downloads/Clef2013/train_2x2_CIELab_256.txt'
testPath = '/home/leonidas/Downloads/Clef2013/test_2x2_CIELab_256.txt'
trainLabelPath = '/home/leonidas/Downloads/Clef2013/train_2x2_CIELab_256_labels.txt'
testLabelPath = '/home/leonidas/Downloads/Clef2013/test_2x2_CIELab_256_labels.txt'
trainPath = '/home/leovala/databases/Clef2013/GBoCFeatures/train_2x2_CIELab_256.txt'
testPath = '/home/leovala/databases/Clef2013/GBoCFeatures/test_2x2_CIELab_256.txt'
trainLabelPath = '/home/leovala/databases/Clef2013/GBoCFeatures/train_2x2_CIELab_256_labels.txt'
testLabelPath = '/home/leovala/databases/Clef2013/GBoCFeatures/test_2x2_CIELab_256_labels.txt'
[trainArray, train_labels, testArray, test_labels,ouClasses] = load.loadFeatures(trainPath=trainPath, trainLabels=trainLabelPath, testPath=testPath,
testLabels=testLabelPath);
# resize to 2D array
trainArray=numpy.reshape(trainArray,(trainArray.shape[0], -1))
testArray=numpy.reshape(testArray,(testArray.shape[0], -1 ))
print "training"
# Random SEARCH ( 0 , 50 ):
# Best parameters set found on development set:
# {'C': 21, 'gamma': 46}
#
print 'tuning now...'
[bestC , bestGamma]= SVM_Tuning.svm_tuning(trainArray,train_labels,testArray,test_labels,randomized=False)
print bestC
print bestGamma
model = train(trainArray,train_labels,c=bestC,g=bestGamma,k='rbf')
# model = train(trainArray,train_labels)
print "testing"
predictions = predict(testArray,model)
accuracy = measures.accuracy(test_labels,predictions)
print accuracy
# Learning Curves
learningCurves.plot_learning_curve(features_train=trainArray, labels_train=train_labels, features_test=testArray, labels_test=test_labels
,K='rbf',C=bestC,G=bestGamma)
def main():
import sys
sys.path.append('../Utils')
sys.path.append('../Utils/Evaluation')
import numpy
from Evaluation import measures
import Load as load
from Tuning import Log_Regression_Tuning
import learningCurves
trainPath = '/home/leonidas/Downloads/Clef2013/train_2x2_CIELab_256.txt'
testPath = '/home/leonidas/Downloads/Clef2013/test_2x2_CIELab_256.txt'
trainLabelPath = '/home/leonidas/Downloads/Clef2013/train_2x2_CIELab_256_labels.txt'
testLabelPath = '/home/leonidas/Downloads/Clef2013/test_2x2_CIELab_256_labels.txt'
[trainArray, train_labels, testArray, test_labels,outputClasses] = load.loadFeatures(trainPath=trainPath, trainLabels=trainLabelPath, testPath=testPath,
testLabels=testLabelPath);
# resize to 2D array
trainArray=numpy.reshape(trainArray,(trainArray.shape[0], -1 ))
testArray=numpy.reshape(testArray,(testArray.shape[0], -1 ))
[bestC , bestAccuracy] = Log_Regression_Tuning.findBestCost(features_train=trainArray, labels_train=train_labels, features_test=testArray, labels_test=test_labels)
print 'best accuracy : {0} , best cost : {1}'.format(bestAccuracy,bestC)
model = train(trainArray,train_labels,c=bestC)
predictions = predict(testArray,model)
print predictions
accuracy = measures.accuracy(test_labels,predictions)
print ' acc : {0}'.format(accuracy)
learningCurves.plot_learning_curve(features_train=trainArray, labels_train=train_labels, features_test=testArray, labels_test=test_labels,classifier="LR",C=bestC)
def ManyClassifiers():
import sys
sys.path.append('../Utils')
sys.path.append('../Utils/Evaluation')
import Load as load
from Evaluation import measures
from Utils.Tuning import SVM_Tuning
# trainFolder = '/home/leonidas/Desktop/images/TrainSet'
# testFolder = '/home/leonidas/Desktop/images/TestSet'
#
# trainFolder = '/home/leonidas/Desktop/images/SampleImages - Copy - Copy'
# testFolder = '/home/leonidas/Desktop/images/SampleImages - Copy'
#
#
# [trainArray, train_labels, testArray, test_labels,outputClasses] = matrix.load_dataset(trainFolder=trainFolder,testFolder=testFolder
# ,imageSize=(250,250))
# print (trainArray.shape)
# print (testArray.shape)
trainPath = '/home/leonidas/Downloads/Clef2013/train_2x2_CIELab_256.txt'
testPath = '/home/leonidas/Downloads/Clef2013/test_2x2_CIELab_256.txt'
trainLabelPath = '/home/leonidas/Downloads/Clef2013/train_2x2_CIELab_256_labels.txt'
testLabelPath = '/home/leonidas/Downloads/Clef2013/test_2x2_CIELab_256_labels.txt'
[trainArray, train_labels, testArray, test_labels,outputClasses] = load.loadFeatures(trainPath=trainPath, trainLabels=trainLabelPath, testPath=testPath,
testLabels=testLabelPath);
import numpy
# resize to 2D array
trainArray=numpy.reshape(trainArray,(trainArray.shape[0], -1))
testArray=numpy.reshape(testArray,(testArray.shape[0], -1))
print (trainArray.shape)
print (testArray.shape)
print outputClasses
print ("Training")
# train classifiers
classifiers = []
for i in range(0,outputClasses):
temp_labels = [1 if x == i else 0 for x in train_labels]
# print ('mpika tuning for classifier : {0}'.format(i))
# [bestC , bestGamma]= SVM_Tuning.svm_tuning(trainArray,temp_labels,testArray,test_labels,randomized=False,kernel='linear')
model = train(trainArray,temp_labels,g=16,c=16,k='linear')
classifiers.append(model)
print ("Testing")
classifiers_predictions = []
for model in classifiers:
predictions = predictProba(testArray,model)
classifiers_predictions.append(predictions)
# classifiers_predictions = numpy.asarray(classifiers_predictions)
# take maximum class probability from classifiers
# predicted_labels = numpy.argmax(classifiers_predictions,axis=0)
# print predicted_labels
accuracy = measures.accuracy(test_labels,numpy.argmax(classifiers_predictions,axis=0) )
print 'Accuracy with many classifiers : {0} '.format(accuracy)
