zzeitt / forFontAugmentation

README

Environment

• python 3.6.2

conda create -n Datasets python=3.6

• packages

pip install -r requirements.txt

How to run

There are two steps for image augmentation. The first step is applying shading effect by replacing background with gradient image and adjusting intensity of foreground. The second step is applying common augmentations, including perspective transform, rotation, Gaussian blur and Gaussian noise.

Applied augmentations:

1. background replacement
2. foreground adjustment
3. perspective transform
4. rotation
5. Gaussian blur
6. Gaussian noise

Run the codes by calling aug_image() in augmentation.py.

import cv2
from augmentation import aug_image

img_src = cv2.imread('./img_src.png')
img_out = aug_image(img_src)

Examples

Before	After