run.py

import numpy as np
import os
from data_helper import Data
from keras.models import Sequential, load_model
from keras.layers import Dense, Activation
from keras.optimizers import SGD
from scipy import misc


def save_model():
    model.save('neural_net.h5')


data = Data()
data.load_data()

training_data = data.training_input
training_labels = data.training_labels
test_data = data.test_input
test_labels = data.test_labels

model = Sequential()

if os.path.isfile('neural_net.h5'):
<<<<<<< Updated upstream
    load_choice = input(
        'Do you want to load the previously saved neural network?[Y/n]')
    if load_choice.lower() == 'y':
        model = load_model('neural_net.h5')
    else:
        sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
=======
	load_choice = input('Do you want to load the previously saved neural network?[Y/n]')
	if load_choice.lower() == 'y':
	    model = load_model('neural_net.h5')
	else:
	    sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
>>>>>>> Stashed changes

        model.add(Dense(30, activation='sigmoid', input_shape=(16384,)))
        model.add(Dense(62, activation='sigmoid'))
        model.compile(loss='mean_squared_error', optimizer=sgd)

        print('Starting to train...')

        history = model.fit(training_data, training_labels,
                            epochs=20, batch_size=10)
        score = model.evaluate(test_data, test_labels, batch_size=32)

        print('\nTest accuracy: {}'.format(score))

<<<<<<< Updated upstream
        save_choice = input('Do you want to save this model? [Y/n]')
        if save_choice.lower() == 'y':
            save_model()
=======
	    save_choice = input('Do you want to save this model? [Y/n]')
	    if save_choice.lower() == 'y':
	        save_model()
>>>>>>> Stashed changes

image = misc.imread('10.jpg', mode='L').reshape(1, 16384)
classes = model.predict(image)
print(np.argmax(classes))