loadData.py

#!/usr/bin/env python

"""
The main file for load models 

author: Xiaowei Huang
"""

import sys
sys.path.append('networks')
sys.path.append('safety_check')
sys.path.append('adversary_generation')

import time
import numpy as np
import scipy.io as sio
import matplotlib.pyplot as plt
from pylab import *

# keras
from keras.models import Model, Sequential
from keras.layers import Input, Dense
import keras.optimizers


# visualisation
from keras.utils.visualize_util import plot
#
from keras.datasets import mnist
from keras.utils import np_utils

# for training cifar10
from keras.preprocessing.image import ImageDataGenerator




from configuration import *


# training the model from data
# or read the model from saved data file
# then start analyse the model 
def loadData():
    
    # construct model according to the flage
    # whichMode == "read" read from saved file 
    # whichMode == "train" training from the beginning 
    if whichMode == "train" and dataset == "mnist": 
    
        (X_train, Y_train, X_test, Y_test, batch_size, nb_epoch) = NN.read_mnist_dataset()
    
        print "Building network model ......"
        model = NN.build_mnist_model()

        start_time = time.time()
        model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch,
                  verbose=1, validation_data=(X_test, Y_test))
        score = model.evaluate(X_test, Y_test, verbose=0)
        print('Test score:', score[0])
        print('Test accuracy:', score[1])
        print("Fitting time: --- %s seconds ---" % (time.time() - start_time))   
        print("Training finished!")
    
        # save model
        json_string = model.to_json()
        open('%s/mnist.json'%(directory_model_string), 'w').write(json_string)
        model.save_weights('%a/mnist.h5'%directory_model_string, overwrite=True)
        sio.savemat('%s/mnist.mat'%directory_model_string, {'weights': model.get_weights()})
        print("Model saved!")
        
    elif whichMode == "read" and dataset == "mnist": 
        print("Start loading model ... ")
        model = NN.read_model_from_file('%s/mnist.mat'%directory_model_string,'%s/mnist.json'%directory_model_string)
        print("Model loaded!")
        #test(model)

    elif whichMode == "train" and dataset == "cifar10": 
    
        (X_train,Y_train,X_test,Y_test, img_channels, img_rows, img_cols, batch_size, nb_classes, nb_epoch, data_augmentation) = NN.read_dataset()
        
        print "Building network model ......"
        model = NN.build_model(img_channels, img_rows, img_cols, nb_classes)
        
        X_train = X_train.astype('float32')
        X_test = X_test.astype('float32')
        X_train /= 255
        X_test /= 255

        start_time = time.time()
        if not data_augmentation:
            print('Not using data augmentation.')
            model.fit(X_train, Y_train,
                      batch_size=batch_size,
                      nb_epoch=nb_epoch,
                      validation_data=(X_test, Y_test),
                      shuffle=True)
        else:
            print('Using real-time data augmentation.')

            # this will do preprocessing and realtime data augmentation
            datagen = ImageDataGenerator(
                featurewise_center=False,  # set input mean to 0 over the dataset
                samplewise_center=False,  # set each sample mean to 0
                featurewise_std_normalization=False,  # divide inputs by std of the dataset
                samplewise_std_normalization=False,  # divide each input by its std
                zca_whitening=False,  # apply ZCA whitening
                rotation_range=0,  # randomly rotate images in the range (degrees, 0 to 180)
                width_shift_range=0.1,  # randomly shift images horizontally (fraction of total width)
                height_shift_range=0.1,  # randomly shift images vertically (fraction of total height)
                horizontal_flip=True,  # randomly flip images
                vertical_flip=False)  # randomly flip images

            # compute quantities required for featurewise normalization
            # (std, mean, and principal components if ZCA whitening is applied)
            datagen.fit(X_train)

            # fit the model on the batches generated by datagen.flow()
            model.fit_generator(datagen.flow(X_train, Y_train,
                                batch_size=batch_size),
                                samples_per_epoch=X_train.shape[0],
                                nb_epoch=nb_epoch,
                                validation_data=(X_test, Y_test))
                                
        score = model.evaluate(X_test, Y_test, verbose=0)
        print('Test score:', score[0])
        print('Test accuracy:', score[1])
        print("Fitting time: --- %s seconds ---" % (time.time() - start_time))   
        print("Training finished!")
    
        # save model
        json_string = model.to_json()
        open('%s/cifar10.json'%directory_model_string, 'w').write(json_string)
        model.save_weights('%s/cifar10.h5'%directory_model_string, overwrite=True)
        sio.savemat('%s/cifar10.mat'%directory_model_string, {'weights': model.get_weights()})
        print("Model saved!")
        
    elif whichMode == "read" and dataset == "cifar10": 
        print("Start loading model ... ")
        (X_train,Y_train,X_test,Y_test, img_channels, img_rows, img_cols, batch_size, nb_classes, nb_epoch, data_augmentation) = NN.read_dataset()
        model = NN.read_model_from_file(img_channels, img_rows, img_cols, nb_classes, '%s/cifar10.mat'%directory_model_string,'%s/cifar10.json'%directory_model_string)
        model.compile(optimizer='rmsprop', loss='categorical_crossentropy')
        print("Model loaded!")
        
    elif whichMode == "train" and dataset == "imageNet": 
    
        (img_channels, img_rows, img_cols, batch_size, nb_classes, nb_epoch, data_augmentation) = NN.read_dataset()
        
        print "Building network model ......"
        model = NN.build_model(img_channels, img_rows, img_cols, nb_classes)
    
        # load weights
        model.load_weights('imageNet/imageNet.h5')

        # save model
        json_string = model.to_json()
        open('%s/imageNet.json'%directory_model_string, 'w').write(json_string)
        model.save_weights('%s/imageNet.h5'%directory_model_string, overwrite=True)
        sio.savemat('%s/imageNet.mat'%directory_model_string, {'weights': model.get_weights()})
        print("Model saved!")
        
    elif whichMode == "read" and dataset == "imageNet": 
        print("Start loading model ... ")
        (img_channels, img_rows, img_cols, batch_size, nb_classes, nb_epoch, data_augmentation) = NN.read_dataset()
        model = NN.read_model_from_file(img_channels, img_rows, img_cols, nb_classes, '%s/imageNet.mat'%directory_model_string,'%s/imageNet.json'%directory_model_string)
        model.compile(optimizer='rmsprop', loss='categorical_crossentropy')
        print("Model loaded!")
        
    elif whichMode == "train" and dataset == "twoDcurve": 
        # define and construct model

        # load data
        N_samples = 5000
        N_tests = 1000
        x_train, y_train, x_test, y_test = NN.load_data(N_samples,N_tests)
        
        print "Building network model ......"
        model = NN.build_model()
    
        plot(model, to_file='twoDcurve_pic/model.png')

        # visualisation

        # configure learning process
        sgd = keras.optimizers.SGD(lr=0.001, decay=1e-6, momentum=0.9, nesterov=True)
        model.compile(optimizer=sgd, loss={'output': 'mse'})

        model.summary()

        start_time = time.time()
        model.fit({'data': x_train}, {'output': y_train}, nb_epoch=3000, validation_split=0.1, verbose=0)
        print("Fitting time: --- %s seconds ---" % (time.time() - start_time))
    
        print("Training finished!")

        # save model
        json_string = model.to_json()
        open('%s/MLP.json'%directory_model_string, 'w').write(json_string)
        model.save_weights('%s/MLP.h5'%directory_model_string, overwrite=True)
        sio.savemat('%s/MLP.mat'%directory_model_string, {'weights': model.get_weights()})
        print("Model saved!")

    elif whichMode == "read" and dataset == "twoDcurve": 
        print("Start loading model ... ")
        model = NN_twoDcurve.read_model_from_file('%s/MLP.mat'%directory_model_string,'%s/MLP.json'%directory_model_string)
        #model.summary()
        print("Model loaded!")

    
    return (model)

        
"""
   validate the model by the test data from the package
""" 
def test(model):

    (X_train, y_train), (X_test, y_test) = mnist.load_data()
    X_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols)
    X_test = X_test.astype('float32')
    X_test = X_test.astype('float32')
    X_test /= 255

    Y_test = np_utils.to_categorical(y_test, nb_classes)

    print("Start testing model ... ")
    # prediction after training
    start_time = time.time()
    y_predicted = model.predict(X_test)
    print y_predicted

    print("Testing time: --- %s seconds ---" % (time.time() - start_time))