When you develop a new deep learning model, it should be trained with a huge input dataset. When the input training dataset size is small for computer vision applications, the performance of the model can be increased with Image augmentation.
Image augmentation is process of creating new dataset from the existing images with few image processing operations like
- Resize the image
- Image Blur
- Image transformation: Affine transformation, Gaussian Blur
- Image rotation
- Histogram Equalization
- Image padding
- Hue Image etc