-
Notifications
You must be signed in to change notification settings - Fork 0
/
2021CSB1102_Kartik_PA2.py
158 lines (129 loc) · 5.87 KB
/
2021CSB1102_Kartik_PA2.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
'''
Kartik Tiwari
- Face Morphing program that takes two images as input and generates n frams (specified by user) to perform face morphing
Using delaunay triangulation. Please ensure that the dimensions of both images in input are same.
"shape_predictor_68_face_landmarks.dat" used to detect 68 landmark points on face
'''
from imutils import face_utils #For face points mapping
import dlib #For the model to detect landmarks automatically
# https://www.geeksforgeeks.org/how-to-install-dlib-library-for-python-in-windows-10/
# To install dlib, first pip install cmake, then pip install dlib
import os
import cv2 #OpenCV library
import numpy #Numpy library
import imageio
#To return indices of triangles in image
def find_i(point, tr):
cnt = 0
for index in tr:
if point == index:
return cnt
cnt += 1
#Delaunay triangulations
def triangulate(img, point):
#Subdiv class instance
subs = cv2.Subdiv2D((0, 0, img.shape[1], img.shape[0]))
for p in point:
subs.insert(p)
#Returns triangles vertices
return subs.getTriangleList()
#Combine input and output via affine transform
def combineResult(img1, img2, img, t1, t2, t, alpha):
rect1 = cv2.boundingRect(numpy.float32([t1]))
rect2 = cv2.boundingRect(numpy.float32([t2]))
rect = cv2.boundingRect(numpy.float32([t]))
t1_cropped, t2_cropped, t_cropped = [],[],[]
for i in range(3):
t1_cropped.append(((t1[i][0] - rect1[0]), (t1[i][1] - rect1[1])))
t2_cropped.append(((t2[i][0] - rect2[0]), (t2[i][1] - rect2[1])))
t_cropped.append(((t[i][0] - rect[0]), (t[i][1] - rect[1])))
mask = numpy.zeros((rect[3], rect[2], 3), dtype=numpy.float32)
cv2.fillPoly(mask, [numpy.int32(t_cropped)], (1.0, 1.0, 1.0), 16)
img1_roi = img1[rect1[1]:rect1[1] + rect1[3], rect1[0]:rect1[0] + rect1[2]]
img2_roi = img2[rect2[1]:rect2[1] + rect2[3], rect2[0]:rect2[0] + rect2[2]]
size = (rect[2], rect[3])
warp1 = cv2.getAffineTransform(numpy.float32(t1_cropped), numpy.float32(t_cropped))
warp_img1 = cv2.warpAffine(img1_roi, warp1, size)
warp2 = cv2.getAffineTransform(numpy.float32(t2_cropped), numpy.float32(t_cropped))
warp_img2 = cv2.warpAffine(img2_roi, warp2, size)
result_roi = (1. - alpha) * warp_img1 + alpha * warp_img2
img[rect[1]:rect[1]+rect[3], rect[0]:rect[0]+rect[2]] = img[rect[1]:rect[1]+rect[3], rect[0]:rect[0]+rect[2]] * (1-mask) + result_roi * mask
#Mark the landmarks and return coordinates
def get_points(img):
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
rects = detector(gray, 0)
for (i, rect) in enumerate(rects):
shaped = predictor(gray, rect)
shaped = face_utils.shape_to_np(shaped)
h, w, c = img.shape
triangle = []
for coord in shaped:
triangle.append((int(coord[0]),int(coord[1])))
triangle.extend([(0,0), (0,h-1), (w-1,h-1), (w-1,0)])
return triangle
if __name__ == '__main__':
frames = int(input("How many frames to generate?")) #20 recommended (optimal)
fps = frames/2
#Open the model to detect face
p = "shape_predictor_68_face_landmarks.dat"
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(p)
#Path of the two images to be morphed
image = cv2.imread('a1.jpg')
image1 = cv2.imread('b1.jpg')
triangle1 = []
triangle2 = []
print("Get tiepoint manually (press 1) or automatically (press 2)")
n = int(input())
if(n==1):
with open("file1.txt", "r") as f:
for i in f:
a, b, c, d = map(int, i.split())
triangle1.append((a, b))
triangle2.append((c, d))
if(n==2):
triangle1 = get_points(image)
triangle2 = get_points(image1)
dlny1 = triangulate(image, triangle1)
triangle_index = []
for t in dlny1:
pt1, pt2, pt3 = (t[0], t[1]), (t[2], t[3]), (t[4], t[5])
# Creating a combined point of (x,y)
add = [(find_i(pt1, triangle1), find_i(pt2, triangle1), find_i(pt3, triangle1))]
triangle_index.extend(add)
#Create a video from the generated frames
fourcc = cv2.VideoWriter_fourcc(*"mp4v")
h, w, c = image.shape
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
height, width, channels = image.shape
fps = frames
parent_dir = './'
v_dir = 'output_gif'
img_dir = 'output_images'
path = os.path.join(parent_dir, v_dir)
os.mkdir(path)
path = os.path.join(parent_dir, img_dir)
os.mkdir(path)
output_path = './output_images'
gif_path = './output_gif/morph.gif'
with imageio.get_writer(gif_path, mode='I', duration=0.05) as writer:
for frame in range(frames):
alpha_factor = (frame + 1) / frames
triangle_middle = []
for i in range(len(triangle1)):
x = int(((1-alpha_factor) * triangle1[i][0]) + (alpha_factor * triangle2[i][0]))
y = int(((1-alpha_factor) * triangle1[i][1]) + (alpha_factor * triangle2[i][1]))
triangle_middle.append((x, y))
morphed_image = numpy.zeros(image.shape, dtype=image.dtype)
for j in range(len(triangle_index)):
x, y, z = triangle_index[j][0], triangle_index[j][1], triangle_index[j][2]
t1 = [triangle1[x], triangle1[y], triangle1[z]]
t2 = [triangle2[x], triangle2[y], triangle2[z]]
t = [triangle_middle[x], triangle_middle[y], triangle_middle[z]]
combineResult(image, image1, morphed_image, t1, t2, t, alpha_factor)
cv2.imshow('Morphed Face', numpy.uint8(morphed_image))
cv2.imwrite(f'{output_path}/{frame}.jpg', morphed_image)
gif_image = cv2.cvtColor(morphed_image, cv2.COLOR_BGR2RGB)
writer.append_data(gif_image)
cv2.waitKey(50)
cv2.destroyAllWindows()