train.py

from network import CNN
from imageLoader import imageload
import sklearn.utils
import theano
import numpy as np


def get_network(lr=theano.shared(np.cast['float32'](0.1)), load=False):
    E = CNN.EmotionClassifier(epochs=500, learning_rate=lr)
    if load:
        E.load_network_state()
    return E


def get_train_test():
    x, y = imageload.load_ck_set()
    x, y = sklearn.utils.shuffle(x, y, random_state=42)
    split_point = int(x.shape[0] * 0.70)
    return x[:split_point], y[:split_point], x[split_point:], y[:split_point]


learning_rate = theano.shared(np.cast['float32'](0.15))
training_rate = input("\r\nEnter a training learning rate or no value for default\r\n>")
try:
    training_rate = float(training_rate)
    learning_rate = np.cast['float32'](training_rate)
except Exception:
    pass
load_state = input("Do you want to load the saved state? y/n\r\n>")

print("Initializing Network State (from saved: " + str(load_state).lower() + ")")
emotion = get_network(lr=learning_rate, load=str(load_state).lower() == "y")

print("Loading Training Data:")
train_x, train_y, test_x, test_y = get_train_test()

while True:
    enter = input("\r\nPress Enter to Train Network\nType Anything else to skip this round of training\r\n>")
    if str(enter).lower() is not "":
        break
    emotion.train(train_x, train_y, 40)
save = input("Do you want to save the training? y/n\r\n>")
if str(save).lower() == "y":
    emotion.save_network_state()