Data is a very important factor in deep learning. The more data you have, the better your model can perform. If you do not have enough data, it is possible to reproduce them artificially. Using the repository I developed below, you can artificially increase your data by using its features such as vertical mirroring, horizontal mirroring and brightness change.
You can easily augmentation labeled images and txt files in yolo format.
Augmentation types:
- Horizantal Flip
- Vertical Flip
- Horizantal and Vertical Flip
- Random Brightness
The input parameters can be changed using the command line :
yolo_aug.py -i <input_dir> -t <aug_type (hflip,vflip,hvflip,bright)> -e <image extension (jpg,jpeg,png ...)> -o <output_dir>
For running example :
python3 yolo_aug.py -i images -t hflip -e jpg -o aug
Original image  |
Vertical image  |
Horizantal and Vertical image  |
Horizantal image 
|