forked from openai/glow
-
Notifications
You must be signed in to change notification settings - Fork 0
/
graphics.py
executable file
·68 lines (61 loc) · 2.21 KB
/
graphics.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
import numpy as np
from PIL import Image
import time
import threading
def save_image(x, path):
im = Image.fromarray(x)
im.save(path, optimize = True)
return
# Assumes [NCHW] format
def save_raster(x, path, rescale=False, width=None):
t = threading.Thread(target=_save_raster, args=(x, path, rescale, width))
t.start()
def _save_raster(x, path, rescale, width):
x = to_raster(x, rescale, width)
save_image(x, path)
# Shape: (n_patches,rows,columns,channels)
def to_raster_old(x, rescale=False, width=None):
x = np.transpose(x, (0,3,1,2))
#x = x.swapaxes(2, 3)
if len(x.shape) == 3:
x = x.reshape((x.shape[0],1,x.shape[1],x.shape[2]))
if x.shape[1] == 1:
x = np.repeat(x, 3, axis=1)
if rescale:
x = (x - x.min()) / (x.max() - x.min()) * 255.
x = np.clip(x, 0, 255)
assert len(x.shape) == 4
assert x.shape[1] == 3
n_patches = x.shape[0]
if width is None:
width = int(np.ceil(np.sqrt(n_patches))) #result width
height = int(n_patches/width) #result height
tile_height = x.shape[2]
tile_width = x.shape[3]
result = np.zeros((3,int(height*tile_height),int(width*tile_width)), dtype='uint8')
for i in range(height):
for j in range(width):
result[:, i*tile_height:(i+1)*tile_height, j*tile_width:(j+1)*tile_width] = x[i]
return result
# Shape: (n_patches,rows,columns,channels)
def to_raster(x, rescale=False, width=None):
if len(x.shape) == 3:
x = x.reshape((x.shape[0], x.shape[1], x.shape[2], 1))
if x.shape[3] == 1:
x = np.repeat(x, 3, axis=3)
if rescale:
x = (x - x.min()) / (x.max() - x.min()) * 255.
x = np.clip(x, 0, 255)
assert len(x.shape) == 4
assert x.shape[3] == 3
n_batch = x.shape[0]
if width is None:
width = int(np.ceil(np.sqrt(n_batch))) # result width
height = int(n_batch / width) # result height
tile_height = x.shape[1]
tile_width = x.shape[2]
result = np.zeros((int(height * tile_height), int(width * tile_width), 3), dtype='uint8')
for i in range(height):
for j in range(width):
result[i * tile_height:(i + 1) * tile_height, j * tile_width:(j + 1) * tile_width] = x[width*i+j]
return result