Convolutional Neural Network for Object Recognition.py

import numpy as np
from keras.models import Sequential
from keras.layers import Dropout, Dense, Flatten
from keras.optimizers import SGD
from keras.layers.convolutional import Conv2D, MaxPooling2D
from keras.utils import np_utils as u
from keras.datasets import cifar10

#Lets start by loading the Cifar10 data
(X, y), (X_test, y_test) = cifar10.load_data()

#Keep in mind the images are in RGB
#So we can normalise the data by diving by 255
#The data is in integers therefore we need to convert them to float first
X, X_test = X.astype('float32')/255.0, X_test.astype('float32')/255.0

#Then we convert the y values into one-hot vectors
#The cifar10 has only 10 classes, thats is why we specify a one-hot
#vector of width/class 10
y, y_test = u.to_categorical(y, 10), u.to_categorical(y_test, 10)

#Now we can go ahead and create our Convolution model
model = Sequential()
#We want to output 32 features maps. The kernel size is going to be
#3x3 and we specify our input shape to be 32x32 with 3 channels
#Padding=same means we want the same dimensional output as input
#activation specifies the activation function
model.add(Conv2D(32, (3, 3), input_shape=(32, 32, 3), padding='same',
                 activation='relu'))
#20% of the nodes are set to 0
model.add(Dropout(0.2))
#now we add another convolution layer, again with a 3x3 kernel
#This time our padding=valid this means that the output dimension can
#take any form
model.add(Conv2D(32, (3, 3), activation='relu', padding='valid'))
#maxpool with a kernet of 2x2
model.add(MaxPooling2D(pool_size=(2, 2)))
#In a convolution NN, we neet to flatten our data before we can
#input it into the ouput/dense layer
model.add(Flatten())
#Dense layer with 512 hidden units
model.add(Dense(512, activation='relu'))
#this time we set 30% of the nodes to 0 to minimize overfitting
model.add(Dropout(0.3))
#Finally the output dense layer with 10 hidden units corresponding to
#our 10 classe
model.add(Dense(10, activation='softmax'))
#Few simple configurations
model.compile(loss='categorical_crossentropy',
              optimizer=SGD(momentum=0.5, decay=0.0004), metrics=['accuracy'])
#Run the algorithm!
model.fit(X, y, validation_data=(X_test, y_test), epochs=25,
          batch_size=512)
#Save the weights to use for later
model.save_weights("cifar10.hdf5")
#Finally print the accuracy of our model!
print("Accuracy: &2.f%%" %(model.evaluate(X_test, y_test)[1]*100))
#print(model.predict_classes(X[0:1]))
#Prints out a number
#1 - airplane, 2 - automobile, 3 - bird, 4 - cat, 5 - deer, 6 - dog
#7 - frog, 8 - horse, 9 - ship, 10 - truck