image-augs 1.0.1

Image Augs supports Augmentation for Object Detection , Instance Segmentation and classification tasks.

Image Augmentations

This is a bounding box level image augmentation tool, it can perform 14 annotations. The important ones are rotation, affine, zooming in and out, noise, and blur. The augmentations were applied to a fraction of the data (40 - 50 percent of the images out of 100 can be augmented).When rotating or zooming in and out, the bounding box coordinates will also change as the image is rotated or zoomed.

Create a folder first, inside that folder keep your image annotation folder.

Create a virtual environment

pip install virtualenv
#name your environment
python3 -m venv <your env name>

#activate the environment --> for linux user
source <your env name>/bin/activate

#activate the environment  --> for windows user
<your env name>/Source/activate.ps1

Installation (for pip installation)

pip install image_augs

After installation

create a .py script inside your created folder

This Script is for OBJECT DETECTION

#import these modules in your created <scriptname>.py file


from object_detection_aug.combined import main
from object_detection_aug.converter_for_txtToYolo import converter



annotation_folder = 'test_objectDetection'  # YOUR ANNOTATION FOLDER NAME
new_aug_saved_folder = 'test_object'        # AUGMENTATION SAVED FOLDER
resize_image = 640                          # REIZE IMAGE
test_split = 0.20                           # TEST SPLIT 



output = converter(annotation_folder,resize_im=resize_image)

# DO YOUR DESIRED AUGMENTATION
# IF BLUR TRUE AND BLUR_F SET TO 0.7 THEN IT WILL TAKE 50% OF YOUR TRAINNING DATA AND APPLY BLUR ON IT.
dicc = main(folder=new_aug_saved_folder,


    test_split=test_split,

    blurs=True, blur_f=0.7,

    noise=True, noise_f=0.6,

    NB=True, NB_f=0.5,

    hue=True, hue_f=0.5,

    sat=True, sat_f=0.5,

    bright=True, bright_f=0.7,

    contrast=True , contrast_f=0.5,

    rotation=True, rotation_f=0.5,

    zoom=True, zoom_f=0.5,

    affine=True, affine_f=0.5,

    translation=True, translation_f=0.5,

    vertical_flip=True, vertical_f=0.5)

#results will be saved in < your given folder >

This Script is for INSTANCE SEGMENTATION

#import these modules in your created <scriptname>.py file
from instance_seg.json_reader_poly import PolygonAugmentation

annotation_folder = '<YOUR ANNOTATION FOLDER NAME>'    
new_aug_saved_folder = '< AUGMENTATION SAVED FOLDER>'
resize_image = 640
train_split = 0.80

im_aug_helper = PolygonAugmentation(aug_save_folder_name=new_aug_saved_folder,
                                    image_resize=resize_image)

im_aug_helper.Image_augmentation(annotation_folder,
                                 
                                 train_split=train_split,

                                    blur=True,  blur_f=0.8,

                                    rotate=True, rotate_f = 0.8, 

                                    noise=True, noise_f=0.8,

                                    perspective=True, perspective_f = 0.8,

                                    affine=True, affine_f=0.8,

                                    brightness=True, brightness_f=0.8,
                                    
                                    hue=True, hue_f=0.8,

                                    removesaturation=True, removesaturation_f=0.8,

                                    contrast=True, contrast_f=0.8,

                                    upflip=True, upflip_f=0.8,

                                    shear=True , shear_f=0.8, 

                                    rotate90=True, rotate90_f =0.8,

                                    blur_and_noise=True, blur_and_noise_f=0.8,

                                    image_cutout = True, image_cutout_f=0.8,
                                    
                                    mix_aug=True, mix_aug_f=0.8,
                                    
                                    temperature_change=True, temperature_change_f=0.8)
#results will be saved in < your given folder >

This Script is for IMAGE CLASSIFICATION

#import these modules in your created <scriptname>.py file
from classification.classification_combined import ImageAugmentHelper


### PARAMS ###
source_folder = '<source folder>'
aug_saved_folder = '<augmentation saved folder>'
train_split = 0.5
image_height = 512
image_width = 512

classification_aug = ImageAugmentHelper(source_folder=source_folder,
                                        aug_save_folder_name=aug_saved_folder,
                                        train_split=train_split,
                                        height=image_height,
                                        width=image_width)


classification_aug.augmentations(

    save_raw_images=True,

    blur=True, blur_f=1.0,

    noise=True,noise_f=1.0,

    horizontalFlip=True, horizontalFlip_f=1.0,

    brightness=True, brightness_f=1.0,

    contrast=True, contrast_f=1.0,

    hue=True, hue_f=1.0,

    saturation=True, saturation_f=1.0,

    zoom=True, zoom_f=1.0,

    perspective=True, perspective_f=1.0,

    translation=True, translation_f=1.0,

    sharpen=True, sharpen_f=1.0,
    
    randomShadow=True, randomShadow_f=1.0
)


Use github to clone [image_augmentations](https://github.com/Souviksaha1998/Image_augmentations) repo.




## Contributing
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

## License
[MIT](https://choosealicense.com/licenses/mit/)