akhilgakhar / mosaic_data_augmentation_yolo

Implement mosaic image augmentation with YOLO format

This is an implementation for mosaic image augmention with YOLO format

Run code to perform mosaic augmentation:

python main.py --width 800 --height 800 --scale_x 0.4 --scale_y 0.6 --min_area 500 --min_vi 0.3

You can change parameters to fit with your dataset. To know details about arguments in the command above, run

python main.py --help

• --width: width of mosaic-augmented image
• --height: height of mosaic-augmented image
• --scale_x: scale_x - scale by width => define width of the top left image
• --scale_y: scale_y - scale by height => define height of the top left image
• --min_area: min area of box after augmentation we will keep
• --min_vi: min area ratio of box after/before augmentation we will keep

Example

With 4 input images

Image 1	Image 2	Image 3	Image 4

Output mosaic augmented image with YOLO format

Mosaic-augmented image	Image with bboxes

Notes

• Original images in the folder images
• Orginal labels (bounding boxes) in the folder labels
• Image and label have the same name. For example: img_0.jpeg - img_0.txt.
• Format of image can be: .jpg, .jpeg, .png ...
• Mosaic-augmented images will be in the folder augmentation/mosaic_images. New labels will be in the folder augmentation/mosaic_labels. These folder will be automatically created when you run the code above.