import numpy as np
from scipy import misc
import pandas as pd
import matplotlib.pyplot as plt
from PIL import Image

# Create first network with Keras
from keras.models import Sequential
from keras.layers import Dense
import numpy
from keras.models import load_model

# fix random seed for reproducibility
seed = 7
numpy.random.seed(seed)

#loading the model from h5 file
model = load_model('my_model.h5')

# load the data which can be tested
#Converting the mnist test set into an image
test_images = pd.read_csv('test.csv').values
dir = "images/"
aList = ['images/test0.png','images/test1.png','images/test2.png','images/test4.png','images/test5.png','images/test6.png']
for i in range(6):
	#img_var = np.reshape(test_images[i], (-1, 28))
	#print np.shape(img_var)
	#img_var = img_var.astype(np.uint8)
	#img = Image.fromarray(img_var) 
	#img.save(dir + 'test'+str(i)+'.png')

	#read an image from the folder
	imgOld = misc.imread(aList[i],flatten=True, mode='L')    
	#imgOld = misc.imread(dir + 'test'+str(i)+'.png',flatten=True, mode='L') 
	img = misc.imresize(imgOld, (28, 28))
	test_img = np.array([img])
	#print test_img.shape
		
	X_test = test_img.reshape(test_img.shape[0], 1, 28, 28).astype('float32')
	X = X_test / 255
	print ('------------------------------------------------------------------------\n')	
	#print X

	# calculate predictions using the KERAS MODEL!!!!!
	predictions = model.predict(X)
	#print('predicted_lables({0})'.format(predictions))
	print ('Predicted_Digit({0})'.format(np.argmax(predictions)))

	"""
	#read an image from the folder
	imgOld = misc.imread(dir + 'test'+str(i)+'.png',flatten=True)      
	img = misc.imresize(imgOld, (28, 28))
	test_img = np.array(img)
	test_img = test_img.astype(np.float)
	# convert from [0:255] => [0.0:1.0]
	test_images = np.multiply(test_img, 1.0 / 255.0)
	print('test_images({0[0]},{0[1]})'.format(test_images.shape))
	test_images_f= np.ravel(test_images)
	test_images_n = np.reshape(test_images_f, (1,784))
	print ('-d	------------------------------------------------------------------------\n')	
	#print (test_images_n)
	# predict test set
	predicted_lables = predict.eval(feed_dict={x: test_images_n, keep_prob: 1.0})
	print('predicted_lables({0})'.format(predicted_lables))
	"""
	# round predictions
	#rounded = [round(x) for x in predictions]
	#print(rounded)
	#print('predicted_lables({0})'.format(predictions))