Homogprahies useful for data augmentation. Quite fast.

homography.py

import numpy as np
from skimage import io
from skimage import transform
import matplotlib.pyplot as plt
import math
import time

I = io.imread('random_img.jpg')

# Homography
theta = -10*np.pi/180
rotation_matrix = np.array([[math.cos(theta), -math.sin(theta), 0],
                     [math.sin(theta),  math.cos(theta), 0],
                     [0,            0,         1]])

shear = 0.1
shear_matrix = np.array([[1.0, -math.sin(shear), 0.0],
                            [0.0, math.cos(shear), 0.0],
                            [0.0, 0.0, 1.0]])

tx, ty = 10,20
translation_matrix = np.array([[1, 0, tx],
                               [0, 1, ty],
                               [0, 0, 1 ]])

H = np.dot(np.dot(rotation_matrix, translation_matrix), shear_matrix)

# Do all of these at once
t0 = time.time()
t = transform.ProjectiveTransform(H)
I1 = transform.warp(I, t, mode='edge')
t1 = time.time()

print('time taken all at once: ', t1-t0)

plt.imshow(I1)
plt.show()


# Do these one after another 
t0 = time.time()
t = transform.ProjectiveTransform(rotation_matrix)
I1 = transform.warp(I, t, mode='edge')
t = transform.ProjectiveTransform(translation_matrix)
I1 = transform.warp(I1, t, mode='edge')
t = transform.ProjectiveTransform(shear_matrix)
I1 = transform.warp(I1, t, mode='edge')
t1 = time.time()
print('time taken one after another: ', t1-t0)


plt.imshow(I1)
plt.show()

This code creates an Affine transformation and not Homography