ImageDataGenerator (in-place augmentation)

inplaceAugmentation.py

import tensorflow 
from tensorflow.keras.preprocessing.image import ImageDataGenerator

# Create training ImageDataGenerator object
train_data_gen = ImageDataGenerator(rotation_range=50,
                                    width_shift_range=0.2,
                                    height_shift_range=0.2,
                                    zoom_range=0.3,
                                    horizontal_flip=True,
                                    vertical_flip=True,
                                    fill_mode='constant',
                                    cval=0,
                                    rescale=1./255)
# Create validation ImageDataGenerator objects				    
valid_data_gen = ImageDataGenerator(rotation_range=45,
                                    width_shift_range=0.2,
                                    height_shift_range=0.2,
                                    zoom_range=0.3,
                                    horizontal_flip=True,
                                    vertical_flip=True,
                                    fill_mode='constant',
                                    cval=0,
                                    rescale=1./255)
test_data_gen = ImageDataGenerator(rescale=1./255)

SEED = 1234
tensorflow.random.set_seed(SEED)
# Training
training_dir = os.path.join(dataset_dir, 'training')
train_gen = train_data_gen.flow_from_directory(training_dir,
                                               target_size=(256, 256),
                                               batch_size=Batch_size,
                                               classes=classes,
                                               class_mode='categorical',
                                               shuffle=True,
                                               seed=SEED)  # targets are directly converted into one-hot vectors

# Validation
valid_dir = os.path.join(dataset_dir, 'valid')
valid_gen = valid_data_gen.flow_from_directory(valid_dir,
                                           target_size=(256, 256),
                                           batch_size=Batch_size, 
                                           classes=classes,
                                           class_mode='categorical',
                                           shuffle=False,
                                           seed=SEED)
# Test
test_dir = os.path.join(dataset_dir, 'testing')
test_gen = test_data_gen.flow_from_directory(test_dir,
                                             target_size=(256, 256),
                                             batch_size=10, 
                                             shuffle=False,
                                             seed=SEED,
                                             class_mode=None,
                                             )