Data Augmentation for Deep Learning

Automatic Data Augmentation for Deep Learning techniques.

Goal

You can generate new data to train neural networks. This is an easy way to prevent the model overfitting.

Features

• Crop
• Rotate
• Noise (Gaussian Noise added)
• Flip-up
• Change brightness

Examples

Original	Crop	Rotate

Noise	Flip-up	Change brightness

Requirements

Software	Version	Required
Python	>= 3.5	Yes
Numpy	>= 1.13	Yes
opencv-python	>= 3.4.2.17	Yes
os	-	Yes
logging	-	Yes
imageio	>= 2.9.0	Yes

Credits

Thanks to Alex Turner