gender.py

from __future__ import print_function

import numpy as np
import matplotlib.pyplot as plt
import keras
from keras.datasets import cifar10
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import BatchNormalization as BN
from keras.layers import GaussianNoise as GN
from keras.optimizers import SGD

from keras.callbacks import LearningRateScheduler as LRS
from keras.preprocessing.image import ImageDataGenerator
import os


os.system('wget https://www.dropbox.com/s/zcwlujrtz3izcw8/gender.tgz')
os.system('tar xvzf gender.tgz')



# Load
x_train = np.load('x_train.npy')
x_test = np.load('x_test.npy')

y_train = np.load('y_train.npy')
y_test = np.load('y_test.npy')

# Stats
print(x_train.shape)
print(y_train.shape)
print(sum(y_train == 1))
print(sum(y_train == 0))
print("{:.4f}".format(sum(y_train == 1)/y_train.shape[0]))
print(x_test.shape)
print(y_test.shape)
print(sum(y_test == 1))
print(sum(y_test == 0))
print("{:.4f}".format(sum(y_test == 1)/y_test.shape[0]))


## View some images
plt.imshow(x_train[0,:,:,: ] )
plt.show()


## Transforms
x_train = x_train.astype('float32')
x_test = x_test.astype('float32')

y_train = y_train.astype('float32')
y_test = y_test.astype('float32')


x_train /= 255
x_test /= 255

print(x_train.shape)
print(x_test.shape)


num_classes = 2

y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)