-
Notifications
You must be signed in to change notification settings - Fork 0
/
seamless.py
81 lines (63 loc) · 1.86 KB
/
seamless.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
from painter import ModelBuilder
from keras import backend as K
from PIL import Image
import numpy as np
print("Imported all")
IMAGE_SIZE = [240, 320] #HW
NUM_LAYERS = 3
NUM_HIDDEN = 20
STEP_COUNT = 100
RADIUS = 0.7
# Activation functions that can be called like f(x)
activations = [
K.softplus,
K.softsign,
K.tanh,
K.sigmoid,
K.hard_sigmoid,
K.sin,
K.cos,
K.abs,
K.log,
K.square,
K.softmax,
K.sqrt,
]
builder = ModelBuilder(activations)
colors = np.zeros((IMAGE_SIZE[0], IMAGE_SIZE[1], 3), dtype=np.float32)
model = None
def make_model():
global model
model = builder.build(NUM_LAYERS, NUM_HIDDEN, 4)
print("Built model")
make_model()
iter = 0
while True:
# The "virtual" time coordinates, go around in a circle over time
angle = 2 * np.pi * iter / STEP_COUNT
coords_u = [RADIUS * np.cos(angle)] * IMAGE_SIZE[1]
coords_v = [RADIUS * np.sin(angle)] * IMAGE_SIZE[1]
for i in range(IMAGE_SIZE[0]):
# The imagespace coordintes
coords_x = [2 * (i / IMAGE_SIZE[0] - 0.5)] * IMAGE_SIZE[1]
coords_y = [2 * (j / IMAGE_SIZE[1] - 0.5) for j in range(IMAGE_SIZE[1])]
coords = np.array([coords_x, coords_y, coords_u, coords_v], dtype=np.float32).T
# Get IMAGE_SIZEx3
colors[i] = model(coords)
# Normalize the output to [0, 255]
data = (255 * (colors - np.min(colors)) / (np.max(colors) - np.min(colors))).astype(np.uint8)
# Check that the image is not a constant color
if np.min(data) != np.max(data) or iter > 0:
img = Image.fromarray(data, "RGB")
img.save("seamless_%d.png" % iter)
iter += 1
print("Iter", iter, "/", STEP_COUNT)
if iter >= STEP_COUNT:
break
else:
print("Bad image")
if iter == 0:
make_model()
else:
print("Failed")
exit()