Data augmentation: ImageDataGenerator vs openCV

Question

I would like to increase the data in my dataset to create a CNN deep learning classification model.

Which is better for the model, using data augmentation by ImageDataGenerator or using openCV to increase the data?

By the way, I am using Keras and floydhub.

Answer 1

An example with library imgaug, Keras, ImageDataGenerator and flow_from_dataframe:

import imgaug as ia
import imgaug.augmenters as iaa

seq = iaa.Sequential([
        iaa.Crop(px=(0, 16)), 
        # crop images from each side by 0 to 16px (randomly chosen)
        iaa.Fliplr(0.5), 
        # horizontally flip 50% of the images
        iaa.GaussianBlur(sigma=(0, 3.0)) 
        # blur images with a sigma of 0 to 3.0
    ])

def augment(img):
        seq_det = seq.to_deterministic()
        aug_image = seq_det.augment_image(img)

        return applications.inception_resnet_v2.preprocess_input(aug_image)

train_generator = image.ImageDataGenerator(preprocessing_function=augment)

train_flow = train_generator.flow_from_dataframe(
        dataframe=train_df,
        directory=train_data_dir,
        x_col="path",
        y_col=columns,
        batch_size=batch_size,
        class_mode="other",
        target_size=(img_height ,img_width),
        shuffle=True
    )

Answer 2

Keras' ImageDataGenerator doesn't offer much support by itself for data augmentation. However it has a parameter called preprocessing_function which allows you to use custom augmentors with it.

I personally use imgaug which offers virtually any augmentation you can think of and works well with ImageDataGenerator like I said.