Python 3: Numpy Array from Generator

console_output

$ python3 numpy_array_timing.py
Using TensorFlow backend.
Approach 1 took 43.884194135665894 s
Approach 2 took 43.12157201766968 s

$ python3 numpy_array_timing.py
Using TensorFlow backend.
Approach 1 took 44.887431383132935 s
Approach 2 took 49.045594930648804 s

$ python3 numpy_array_timing.py
Using TensorFlow backend.
Approach 1 took 46.91734313964844 s
Approach 2 took 52.51308083534241 s

numpy_array_timing.py

# Setup of this script comes from
# https://github.com/dribnet/lfw_fuel/blob/master/example/run-lfw.py

from lfw_fuel import lfw

np.random.seed(1337)  # for reproducibility

from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Activation, Flatten
from keras.layers import Conv2D, MaxPooling2D

from keras.utils import np_utils
from scipy.misc import imresize

batch_size = 128
nb_classes = 2
nb_epoch = 12
feature_width = 32
feature_height = 32

def cropImage(im):
    im2 = np.dstack(im).astype(np.uint8)
    # return centered 128x128 from original 250x250 (40% of area)
    newim = im2[61:189, 61:189]
    sized1 = imresize(newim[:,:,0:3], (feature_width, feature_height), interp="bicubic", mode="RGB")
    sized2 = imresize(newim[:,:,3:6], (feature_width, feature_height), interp="bicubic", mode="RGB")
    return np.asarray([sized1[:,:,0], sized1[:,:,1], sized1[:,:,2], sized2[:,:,0], sized2[:,:,1], sized2[:,:,2]])

# the data, shuffled and split between train and test sets
(X_train, y_train), (X_test, y_test) = lfw.load_data("deepfunneled")



############## Timing Comparison ################

import time

# Approach #1: Create an empty array and populate it
t0 = time.time()
temp = np.empty(np.shape(X_train))
for i, el in enumerate(X_train):
    temp[i] = el
t1 = time.time()
print("Approach 1 took {0} s".format(t1-t0))

# Approach #2: Convert the map to a list, then the list to an array
t2 = time.time()
temp2 = np.asarray(list(map(cropImage,X_train)))
t3 = time.time()
print("Approach 2 took {0} s".format(t3-t2))