import pandas as pd
import confusion_mat as cfm
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report
import numpy as np
import matplotlib.pyplot as plt
from sklearn.neighbors import KNeighborsClassifier
data = pd.read_pickle('Data_KNN.pkl')
labels = pd.read_pickle('Labels_KNN.pkl')
trainX, testX, trainY, testY = train_test_split(data, labels, test_size=.25, random_state=1)
model = KNeighborsClassifier(n_neighbors=200)
H = model.fit(trainX, trainY)
# acc=model.score(testX,testY)
# print("[INFO] raw pixel accuracy: {:.2f}%".format(acc * 100))
predictions = model.predict(testX)
print(classification_report(testY, predictions))
np.set_printoptions(precision=1)
classLabels = ['zero', 'one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight', 'nine']
# Plot non-normalized confusion matrix
cfm.plot_confusion_matrix(testY, predictions, classes=classLabels)
plt.show()
print "%s: %.3f" % (kernel, pred_lin)
accur_lin.append(pred_lin) #Store accuracy in a list
#print "training_labels", training_labels
#print "prediction_labels", prediction_labels
if i == 0:
y_pred = clf.predict(prediction_data)
y_test = prediction_labels
# Compute confusion matrix
class_names = emotions
cnf_matrix = confusion_matrix(y_test, y_pred)
np.set_printoptions(precision=2)
print y_pred
# Plot non-normalized confusion matrix
plt.figure()
plot_confusion_matrix(cnf_matrix,
classes=class_names,
title='Confusion matrix')
plt.tight_layout()
plt.savefig(kernel + '.png')
#plt.show()
print "Mean value %s svm: %.3f" % (kernel, np.mean(accur_lin)
) #Get mean accuracy of the 10 runs
test_on = 'happy'
output = []
# testing on external databases
for i in range(0, 1):
image = cv2.imread('test/' + test_on + '/test' + str(i) + '.png')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
clahe_image = clahe.apply(gray)
x = np.vstack(x)
y = [d['targets'] for d in test_data]
y = np.hstack(y).T
predicted = p.argmax(axis=1)
label = y.argmax(axis=1)
# combine into phenemse
predicted, phoneme_list = combinePhonemes(predicted, statlist)
label, _ = combinePhonemes(label, statlist)
grouped_label = groupedList(label)
grouped_predicted = groupedList(predicted)
accuracy = editdistance.eval(grouped_label, grouped_predicted) / len(grouped_label)
print('accuracy %0.2f%%' % accuracy)
cm = get_confusion_matrix(predicted, label)
# manually normalize sil_0 and sil_1
# cm[39, 39] = 0
# cm[40, 40] = 0
print('calculating accuracy')
accuracy = 1 - np.count_nonzero(predicted - label) / predicted.shape[0]
print(accuracy * 100, '%')
cm[13,13] = 0
plot_confusion_matrix(cm, phoneme_list, title='Phoneme state confusion mat, no silence',save=True)
# model = keras.models.load_model('h3_adagrad_relu_u256_e100.h5py')
# print('predicting..')
# p = model.predict(x)
print('hello world')
NUM_STACK = 5
statlist = pickle.load(open('stateList.pkl', 'rb'))
p = np.load('predicted_test.npy')
train, validation = get_training_and_validation_sets(
np.load('G:/train_data.npz')['data'])
_, __, test_data = standardize_per_training_set(
train, [],
np.load('G:/test_data.npz')['data'])
print('processing data for input')
x = [lmfcc_stack(d['lmfcc'], NUM_STACK) for d in test_data]
x = np.vstack(x)
y = [d['targets'] for d in test_data]
y = np.hstack(y).T
predicted = p.argmax(axis=1)
label = y.argmax(axis=1)
cm = get_confusion_matrix(predicted, label)
# manually normalize sil_0 and sil_1
cm[39, 39] = 0
cm[40, 40] = 0
plot_confusion_matrix(cm, statlist)
# model = keras.models.load_model('h3_adagrad_relu_u256_e100.h5py')
# print('predicting..')
# p = model.predict(x)
print('hello world')
verbose=1)
model.save('E:/Code/htmljscss/model_32_ostu.hdf5')
print('[INFO] evaluating network...')
predictions = model.predict(testX, batch_size=128)
print(classification_report(testY.argmax(axis=1), predictions.argmax(axis=1)))
np.set_printoptions(precision=1)
classLabels = [
'zero', 'one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight',
'nine'
]
# Plot non-normalized confusion matrix
cfm.plot_confusion_matrix(testY.argmax(axis=1),
predictions.argmax(axis=1),
classes=classLabels)
plt.show()
plt.style.use('ggplot')
plt.figure()
plt.plot(np.arange(0, 10), H.history["accuracy"], label="train_acc")
plt.plot(np.arange(0, 10), H.history['val_accuracy'], label='val_acc')
plt.title('Training Accuracy')
plt.xlabel('Epoch #')
plt.ylabel('Accuracy')
plt.legend()
plt.show()
plt.style.use('ggplot')
plt.figure()