Skip to content

Instantly share code, notes, and snippets.

View virtualdvid's full-sized avatar

David Molina virtualdvid

View GitHub Profile
@virtualdvid
virtualdvid / table_4.csv
Last active January 15, 2019 19:32
Flow from directory with Augmentation: Keras vs Gap
Keras: Flow from directory with Augmentation Gap: Flow from directory with Augmentation
PRE-TIME 0.20845842361450195
TIME 143.19440412521362 10.20444941520691
Memory Used: 0.13 GB 0.01 GB
@virtualdvid
virtualdvid / code_11.py
Created January 15, 2019 19:27
Gap: Flow from directory with Augmentation
print("Gap: Flow from directory with Augmentation")
dataset = Images(augment=['rotate=-30,30'], config=['stream'])
dataset.load('flowers')
dataset.minibatch = 32
train_generator = dataset.minibatch
start_mem = psutil.virtual_memory().used
start_time = time.time()
@virtualdvid
virtualdvid / code_10.py
Created January 15, 2019 19:25
Keras: Flow from directory with Augmentation
print("Keras: Flow from directory with Augmentation")
start_mem = psutil.virtual_memory().used
start_time = time.time()
datagen = ImageDataGenerator(rotation_range=30, rescale=1./255)
train_generator = datagen.flow_from_directory('flowers', target_size=(128,128), batch_size=32)
print("PRE-TIME", time.time() - start_time)
@virtualdvid
virtualdvid / table_3.csv
Created January 15, 2019 19:23
Flow with Augmentation: Keras vs Gap
Keras: Flow with Augmentation Gap: Flow with Augmentation
PRE-TIME 0.4257802963256836 0.0036149024963378906
TIME 54.004902839660645 6.712508916854858
Memory Used: -0.00 GB 0.00 GB
@virtualdvid
virtualdvid / code_9.py
Last active January 15, 2019 19:21
Gap: Flow with Augmentation
print("Gap: Flow with Augmentation")
dataset = Images(augment=['rotate=-30,30'])
dataset.load('flowers')
start_mem = psutil.virtual_memory().used
start_time = time.time()
dataset.minibatch = 32
train_generator = dataset.minibatch
@virtualdvid
virtualdvid / code_8.py
Last active January 15, 2019 19:21
Keras: Flow with Augmentation
print("Keras: Flow with Augmentation")
datagen = ImageDataGenerator(rotation_range=30)
start_time = time.time()
datagen.fit(X_train)
train_generator = datagen.flow(X_train, Y_train, batch_size=32)
print("PRE-TIME (fit)", time.time() - start_time)
@virtualdvid
virtualdvid / table_2.csv
Created January 15, 2019 19:11
flow_from_directory Keras vs Gap
Keras: flow_from_directory Gap: flow from directory
PRE-TIME 0.20801329612731934 0.01245427131652832
TIME 89.3897156715393 6.902476787567139
Memory Used: 0.00 GB -0.00 GB
@virtualdvid
virtualdvid / code_7.py
Last active January 15, 2019 19:21
Gap: flow from directory
print("Gap: flow from directory")
start_mem = psutil.virtual_memory().used
start_time = time.time()
dataset = Images(config=['stream'])
dataset.load('flowers')
dataset.minibatch = 32
train_generator = dataset.minibatch
@virtualdvid
virtualdvid / code_6.py
Last active January 15, 2019 19:20
Keras: flow_from_directory
print("Keras: flow_from_directory")
start_mem = psutil.virtual_memory().used
start_time = time.time()
datagen = ImageDataGenerator(rescale=1./255)
train_generator = datagen.flow_from_directory('flowers', target_size=(128,128), batch_size=32)
print("PRE-TIME", time.time() - start_time)
start_time = time.time()
@virtualdvid
virtualdvid / table_1.csv
Created January 15, 2019 18:51
Flow: Keras vs Gap
Keras: Flow Gap: Flow
PRE-TIME 0.38779211044311523 1.8835067749023438e-05
TIME 1.3482673168182373 0.7792487144470215
Memory Used: 1.79 GB -0.00 GB