def SVM_Classify(trainDataPath, trainLabelPath, testDataPath, testLabelPath, kernelType):
trainData = np.array(utils.loadDataFromFile(trainDataPath))
trainLabels = utils.loadDataFromFile(trainLabelPath)
testData = np.array(utils.loadDataFromFile(testDataPath))
testLabels = utils.loadDataFromFile(testLabelPath)
if kernelType == "HI":
gramMatrix = histogramIntersection(trainData, trainData)
clf = SVC(kernel='precomputed')
clf.fit(gramMatrix, trainLabels)
predictMatrix = histogramIntersection(testData, trainData)
SVMResults = clf.predict(predictMatrix)
correct = sum(1.0 * (SVMResults == testLabels))
accuracy = correct / len(testLabels)
print "SVM (Histogram Intersection): " +str(accuracy)+ " (" +str(int(correct))+ "/" +str(len(testLabels))+ ")"
else:
clf = SVC(kernel = kernelType)
clf.fit(trainData, trainLabels)
SVMResults = clf.predict(testData)
correct = sum(1.0 * (SVMResults == testLabels))
accuracy = correct / len(testLabels)
print "SVM (" +kernelType+"): " +str(accuracy)+ " (" +str(int(correct))+ "/" +str(len(testLabels))+ ")"
def SVM_Classify(trainDataPath, trainLabelPath, testDataPath, testLabelPath,
kernelType):
trainData = np.array(utils.loadDataFromFile(trainDataPath))
trainLabels = utils.loadDataFromFile(trainLabelPath)
testData = np.array(utils.loadDataFromFile(testDataPath))
testLabels = utils.loadDataFromFile(testLabelPath)
if kernelType == "HI":
gramMatrix = histogramIntersection(trainData, trainData)
clf = SVC(kernel='precomputed')
clf.fit(gramMatrix, trainLabels)
predictMatrix = histogramIntersection(testData, trainData)
SVMResults = clf.predict(predictMatrix)
correct = sum(1.0 * (SVMResults == testLabels))
accuracy = correct / len(testLabels)
print "SVM (Histogram Intersection): " + str(accuracy) + " (" + str(
int(correct)) + "/" + str(len(testLabels)) + ")"
else:
clf = SVC(kernel=kernelType)
clf.fit(trainData, trainLabels)
SVMResults = clf.predict(testData)
correct = sum(1.0 * (SVMResults == testLabels))
accuracy = correct / len(testLabels)
print "SVM (" + kernelType + "): " + str(accuracy) + " (" + str(
int(correct)) + "/" + str(len(testLabels)) + ")"
def KNN_Classify(trainDataPath, trainLabelPath, testDataPath, testLabelPath):
trainData = np.array(utils.loadDataFromFile(trainDataPath))
trainLabels = utils.loadDataFromFile(trainLabelPath)
testData = np.array(utils.loadDataFromFile(testDataPath))
testLabels = utils.loadDataFromFile(testLabelPath)
KNN = KNeighborsClassifier()
KNN.fit(trainData, trainLabels)
KNN_testLabels = KNN.predict(testData)
correct = sum(1.0 * (KNN_testLabels == testLabels))
accuracy = correct / len(testLabels)
print "KNN: " +str(accuracy)+ " (" +str(int(correct))+ "/" +str(len(testLabels))+ ")"
def KNN_Classify(trainDataPath, trainLabelPath, testDataPath, testLabelPath):
trainData = np.array(utils.loadDataFromFile(trainDataPath))
trainLabels = utils.loadDataFromFile(trainLabelPath)
testData = np.array(utils.loadDataFromFile(testDataPath))
testLabels = utils.loadDataFromFile(testLabelPath)
KNN = KNeighborsClassifier()
KNN.fit(trainData, trainLabels)
KNN_testLabels = KNN.predict(testData)
correct = sum(1.0 * (KNN_testLabels == testLabels))
accuracy = correct / len(testLabels)
print "KNN: " + str(accuracy) + " (" + str(int(correct)) + "/" + str(
len(testLabels)) + ")"
import utils
from network import NNOneHidden
ARG_TRAINED_MODEL_IN = "model"
argParser = argparse.ArgumentParser()
argParser.add_argument("-m",
"--{}".format(ARG_TRAINED_MODEL_IN),
required=True,
help="path to trained model")
args = vars(argParser.parse_args())
trainedModelFileName = args[ARG_TRAINED_MODEL_IN]
print("Loading model from '{}' ...".format(trainedModelFileName))
W, b = utils.loadDataFromFile(trainedModelFileName)
print("Fitting model...")
eel.init("web")
@eel.expose
def classify_image(X):
pixels = list(map(int, X.split(" ")))
#inputData = np.array([1-x for x in pixels])
inputData = np.array(pixels).reshape(784, 1)
# plt.imshow(inputData.reshape(28, 28), cmap=matplotlib.cm.binary)
# plt.axis("off")
# plt.show()
# utils.display(inputData)
required=True,
help="path to training data")
argParser.add_argument("-s",
"--{}".format(ARG_TRAINED_MODEL_OUT),
required=True,
help="path to save trained model")
argParser.add_argument("-b",
"--{}".format(ARG_BATCH_TRAIN),
required=True,
help="train with batch or not")
args = vars(argParser.parse_args())
trainDataFileName = args[ARG_TRAIN_DATA]
print("Getting training data from '{}' ...".format(trainDataFileName))
X_train, y_train = utils.loadDataFromFile(trainDataFileName)
X_test, y_test = utils.loadDataFromFile("data/mnist.data.test")
print(X_train.shape, y_train.shape)
print("Training model batch={}...".format(args[ARG_BATCH_TRAIN]))
nnet = NNOneHidden()
epochs = 1000
hidden_units = 50
if args[ARG_BATCH_TRAIN] == "true":
W, b, train_history, test_history = nnet.mini_batch_train(
X_train,
y_train,
X_test,
y_test,
epochs=epochs,
import numpy as np
#command line arguments for this script
ARG_TRAINED_MODEL_IN = "model"
ARG_TEST_DATA = "test_data"
argParser = argparse.ArgumentParser()
argParser.add_argument("-m", "--{}".format(ARG_TRAINED_MODEL_IN), required=True, help="path to trained model")
argParser.add_argument("-t", "--{}".format(ARG_TEST_DATA), required=True, help="path to test data")
args = vars(argParser.parse_args())
trainedModelFileName = args[ARG_TRAINED_MODEL_IN]
testDataFileName = args[ARG_TEST_DATA]
print("Loading model from '{}' ...".format(trainedModelFileName))
W, b = utils.loadDataFromFile(trainedModelFileName)
print("Loading test data from '{}' ...".format(testDataFileName))
X_test, y_test = utils.loadDataFromFile(testDataFileName)
print("Test data shapes", X_test.shape, y_test.shape)
print("Fitting model and getting predictions...")
nnet = NNOneHidden()
nnet.fit(W, b)
classification = nnet.classify(X_test)
predictions = np.argmax(classification, axis=0)
labels = np.argmax(y_test, axis=0)
correct_predictions = np.sum(predictions == labels) #compute how many predictions were correct
print("Performance : {}/{} correct".format(correct_predictions, predictions.shape[0]))
