Automatic Image Labeling

 

import Augmentor

p = Augmentor.Pipeline('one')

# p.random_distortion(probability=0.05, grid_width=4, grid_height=4, magnitude=4)

p.rotate(probability=0.3, max_left_rotation=25, max_right_rotation=25)

p.flip_left_right(probability=0.2)

p.flip_top_bottom(probability=0.02)

p.flip_random(probability=0.05)

p.rotate90(probability=0.005)

p.rotate180(probability=0.005)

p.rotate270(probability=0.005)

p.skew_tilt(probability=0.001)

p.skew_left_right(probability=0.005)

p.skew_top_bottom(probability=0.005)

p.skew_corner(probability=0.002)

p.skew(probability=0.002)

p.shear(probability=0.005, max_shear_left=10, max_shear_right=10)

p.sample(100)

https://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html

https://machinelearningmastery.com/how-to-configure-image-data-augmentation-when-training-deep-learning-neural-networks/

https://www.tensorflow.org/tutorials/images/data_augmentation

https://towardsdatascience.com/how-to-augmentate-data-using-keras-38d84bd1c80c

from email.mime import image

import os

from PIL import Image

import cv2

# from PIL import Image

from os import listdir

 

# get the path or directory

folder_dir = "one"

for images in os.listdir(folder_dir):

 

    # check if the image end swith png or jpg or jpeg

    if (images.endswith(".png") or images.endswith(".jpg") or images.endswith(".jpeg")):

        # display

        # print(type(images))

        img = Image.open('one/'+images)

        # print(os.path.join(folder_dir, images))

        # width, height = image.size

        width = img.width

        height = img.height

        name=str(images)

        name=name.rsplit( ".", 1 )[ 0 ]

        # name.partition('.')

        f= open('one/'+name+'.txt',"w+")

        f.write("0 0 0 "+str(height)+" "+str(width))

        f.close

  

# # display width and height

        print("The height of the image is: ", height)

        print("The width of the image is: ", width)