import pandas
import numpy
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Dropout
from keras.layers import Flatten
from keras.layers.convolutional import Convolution2D
from keras.layers.convolutional import MaxPooling2D
from keras.utils import np_utils
from keras import backend as K
K.set_image_dim_ordering('th')

# Constants
IMG_ROWS = IMG_COLUMNS = 28
NUM_EPOCHS = 250
DROPOUT = 0.2
BATCH_SIZE = 50
SEED = 0
numpy.random.seed(SEED);

# Load data
train = pandas.read_csv("../input/train.csv").values
test  = pandas.read_csv("../input/test.csv").values

# Data prep
# Separate data into pixels (X) and class (Y)
trainX = train[:, 1:].reshape(train.shape[0], 1, IMG_ROWS, IMG_COLUMNS)
trainX = trainX.astype(float)
testX = test.reshape(test.shape[0], 1, IMG_ROWS, IMG_COLUMNS)
testX = testX.astype(float)
# Normalise pixel data
trainX /= 255.0
testX  /= 255.0
# One hot encode outputs
trainY = np_utils.to_categorical(train[:,0])
numClasses = trainY.shape[1]

# Build model
model = Sequential()
model.add(Convolution2D(32, 3, 3, activation='relu', border_mode='same',
                        input_shape=(1, IMG_ROWS, IMG_COLUMNS)))
model.add(Convolution2D(32, 3, 3, activation='relu', border_mode='same'))
model.add(MaxPooling2D(strides=(2, 2)))
model.add(Convolution2D(64, 1, 3, activation='relu', border_mode='same'))
model.add(Convolution2D(64, 3, 1, activation='relu', border_mode='same'))
model.add(MaxPooling2D(strides=(2, 2)))
model.add(Convolution2D(128, 1, 1, activation='relu', border_mode='same'))
model.add(Convolution2D(128, 1, 1, activation='relu', border_mode='same'))
model.add(MaxPooling2D(strides=(2, 2)))
model.add(Flatten())
model.add(Dropout(0.2))
model.add(Dense(256, activation='relu'))
model.add(Dense(numClasses, activation='softmax'))
model.summary()

# Compile model
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

# Fit model
model.fit(trainX, trainY, nb_epoch=NUM_EPOCHS, batch_size=BATCH_SIZE, verbose=2)

# Evaluate
yPred = cnn.predict_classes(testX)
print(yPred)

# Save
numpy.savetxt('mnist-cnn.csv', numpy.c_[range(1,len(yPred)+1),yPred], delimiter=',', header = 'ImageId,Label', comments = '', fmt='%d')