import cv2
import numpy as np
import homography
if __name__ == '__main__':
img = cv2.imread('test/image_01.jpeg')
#img = cv2.imread('/Users/tal/Dropbox/School/Y4S1/CSC420/Assignment3/data/11.jpg')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img2 = cv2.imread('DVDcovers/shrek2.jpg')
#img2 = cv2.imread('/Users/tal/Dropbox/School/Y4S1/CSC420/Assignment3/data/toy.jpg')
gray2 = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
sift = cv2.xfeatures2d.SIFT_create()
kp2, tdesc = sift.detectAndCompute(gray, None)
kp1, rdesc = sift.detectAndCompute(gray2, None)
print rdesc.shape
print len(kp1)
print len(kp2)
rfeat = np.array([k.pt for k in kp1])
tfeat = np.array([k.pt for k in kp2])
inliers, hom = homography.estimate_homography(rdesc, rfeat, tdesc, tfeat, num_iters=1000)
rmat, tmat = homography.do_matching(rdesc, tdesc)
print rmat.shape
print inliers
print hom
homography.visualize_transformation(img, hom, img2.shape[0], img2.shape[1], 'tes_visualization.jpeg')
cProfile.run('vt.get_most_similar(tdesc)')
most_similar = vt.get_most_similar(tdesc)
best_im_name = ""
best_im = None
most_in = 0
best_rdesc = None
best_rfeat = None
best_hom = None
print most_similar
for train in most_similar:
train_image = cv2.imread(train_dir + os.sep + train)
gray = cv2.cvtColor(train_image, cv2.COLOR_BGR2GRAY)
kp, rdesc = sift.detectAndCompute(gray, None)
rfeat = np.array([k.pt for k in kp])
# cProfile.run('estimate_homography(rdesc, rfeat, tdesc, tfeat)')
inl, hom = estimate_homography(rdesc, rfeat, tdesc, tfeat)
print inl
if inl > most_in:
most_in = inl
best_hom = hom
best_im_name = train
best_rdesc = rdesc
best_rfeat = rfeat
best_im = train_image
print best_im_name
im = visualize_transformation(test_image, best_hom, best_im.shape[0], best_im.shape[1])
im = cv2.putText(im, best_im_name, (5,100), cv2.FONT_HERSHEY_COMPLEX, 1, (255, 0, 0))
cv2.imwrite('full_vis.jpg', im)
# first_image = cv2.imread('DVDcovers/' + images[0])