Image Augmentation for Deep Learning

Since the amount of feed data have a direct impact on the precision of any Neural Network architectures, augmenting of data may seem to be a very good solution to broaden the amount of input data. This process is an increasingly demand for complex object detection and classification purposes. Since training Deep Neural Networks (DNNs) on more data can result in more precise outcomes and expand the training dataset, the process of augmentation may play an important role to provide better classification model. In this project data augmentation is done with imgaug library in Python-3.

Environment:

• Python v3
• Anaconda
• Jupyter

Supported Augmentation Operations:

• Horizontal and Vertical Shift
• Horizontal and Vertical Flip
• Random Rotation
• Random Brightness Change
• Random Zoom/Scale

References:

1. imgaug library in Python
2. machinelearningmastery.com website article: How to Configure Image Data Augmentation in Keras. (See link)