/
test_ensemble.py
188 lines (152 loc) · 7.76 KB
/
test_ensemble.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
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
from util.visualizer import Visualizer
import util.util as util
from data.data_loader import CreateDataLoader
from options.test_options import TestOptions
import torchvision.transforms as transforms
import time
import os
import numpy as np
from PIL import Image
import torch
from torch.autograd import Variable
from collections import OrderedDict
from models.models import create_model
from util import html
import copy
opt = TestOptions().parse(save=False)
opt.nThreads = 1
opt.batchSize = 1
opt.serial_batches = True
opt.no_flip = True
data_loader = CreateDataLoader(opt)
dataset = data_loader.load_data()
dataset_size = len(data_loader)
def __make_power_2(img, base=256, method=Image.BICUBIC):
ow, oh = img.size
h = int(round(oh / base) * base)
w = int(round(ow / base) * base)
if h == 0:
h = base
if w == 0:
w = base
if (h == oh) and (w == ow):
return img
return img.resize((w, h), method)
model, num_params_G, num_params_D = create_model(opt)
model.eval()
rlt_dir = os.path.join(opt.results_dir, 'test')
util.mkdirs([rlt_dir])
transform_list = []
transform_list += [transforms.ToTensor()]
mskimg_transform = transforms.Compose(transform_list)
transform_list = []
transform_list += [transforms.ToTensor()]
msk_transform = transforms.Compose(transform_list)
start_time = time.time()
for i, data in enumerate(dataset):
with torch.no_grad():
msk_img_path = data['path_mskimg'][0]
filename = os.path.basename(msk_img_path)
msk_path = data['path_msk'][0]
oimg = Image.open(msk_img_path).convert('RGB')
omsk = Image.open(msk_path).convert('L')
ow, oh = oimg.size
###
resized_img = __make_power_2(oimg)
resized_msk = __make_power_2(omsk, method=Image.BILINEAR)
rw, rh = resized_img.size
hori_ver = rw // 256
vert_ver = rh // 256
tmp_img = oimg.resize((256, 256), Image.BICUBIC)
tmp_msk = omsk.resize((256, 256), Image.BICUBIC)
np_tmp_img = np.array(tmp_img, np.uint8)
np_tmp_msk = np.array(tmp_msk, np.uint8)
np_resized_img = np.array(resized_img, np.uint8)
np_resized_msk = np.array(resized_msk, np.uint8)
np_resized_msk = np_resized_msk > 0
np_resized_img[:,:,0] = np_resized_img[:,:,0] * (1 - np_resized_msk) + 255 * np_resized_msk
np_resized_img[:,:,1] = np_resized_img[:,:,1] * (1 - np_resized_msk) + 255 * np_resized_msk
np_resized_img[:,:,2] = np_resized_img[:,:,2] * (1 - np_resized_msk) + 255 * np_resized_msk
np_resized_msk = np_resized_msk * 255
img_arr = []
msk_arr = []
###
for hv in range(hori_ver):
for vv in range(vert_ver):
for i in range(256):
for j in range(256):
np_tmp_img[i, j, 0] = np_resized_img[vv + vert_ver*j, hv + hori_ver*i, 0]
np_tmp_img[i, j, 1] = np_resized_img[vv + vert_ver*j, hv + hori_ver*i, 1]
np_tmp_img[i, j, 2] = np_resized_img[vv + vert_ver*j, hv + hori_ver*i, 2]
np_tmp_msk[i, j] = np_resized_msk[vv + vert_ver*j, hv + hori_ver*i]
img_arr.append(np.copy(np_tmp_img))
msk_arr.append(np.copy(np_tmp_msk))
###
compltd_arr = []
for i in range(len(img_arr)):
img = Image.fromarray(img_arr[i])
msk = Image.fromarray(msk_arr[i])
img_90 = img.rotate(90)
msk_90 = msk.rotate(90)
img_180 = img.rotate(180)
msk_180 = msk.rotate(180)
img_270 = img.rotate(270)
msk_270 = msk.rotate(270)
img_flp = img.transpose(method=Image.FLIP_LEFT_RIGHT)
msk_flp = msk.transpose(method=Image.FLIP_LEFT_RIGHT)
compltd_img, reconst_img, lr_x = model(mskimg_transform(img).unsqueeze(0), msk_transform(msk).unsqueeze(0))
compltd_img_90, reconst_img_90, lr_x_90 = model(mskimg_transform(img_90).unsqueeze(0), msk_transform(msk_90).unsqueeze(0))
compltd_img_180, reconst_img_180, lr_x_180 = model(mskimg_transform(img_180).unsqueeze(0), msk_transform(msk_180).unsqueeze(0))
compltd_img_270, reconst_img_270, lr_x_270 = model(mskimg_transform(img_270).unsqueeze(0), msk_transform(msk_270).unsqueeze(0))
compltd_img_flp, reconst_img_flp, lr_x_flp = model(mskimg_transform(img_flp).unsqueeze(0), msk_transform(msk_flp).unsqueeze(0))
np_compltd_img = util.tensor2im(reconst_img.data[0], normalize=False)
np_compltd_img_90 = util.tensor2im(reconst_img_90.data[0], normalize=False)
np_compltd_img_180 = util.tensor2im(reconst_img_180.data[0], normalize=False)
np_compltd_img_270 = util.tensor2im(reconst_img_270.data[0], normalize=False)
np_compltd_img_flp = util.tensor2im(reconst_img_flp.data[0], normalize=False)
new_img_90 = Image.fromarray(np_compltd_img_90)
new_img_90 = new_img_90.rotate(270)
np_new_img_90 = np.array(new_img_90, np.float)
new_img_180 = Image.fromarray(np_compltd_img_180)
new_img_180 = new_img_180.rotate(180)
np_new_img_180 = np.array(new_img_180, np.float)
new_img_270 = Image.fromarray(np_compltd_img_270)
new_img_270 = new_img_270.rotate(90)
np_new_img_270 = np.array(new_img_270, np.float)
new_img_flp = Image.fromarray(np_compltd_img_flp)
new_img_flp = new_img_flp.transpose(method=Image.FLIP_LEFT_RIGHT)
np_new_img_flp = np.array(new_img_flp, np.float)
np_compltd_img = (np_compltd_img + np_new_img_90 + np_new_img_180 + np_new_img_270 + np_new_img_flp) / 5.0
np_compltd_img = np.array(np.round(np_compltd_img), np.uint8)
final_img = Image.fromarray(np_compltd_img, mode="RGB")
np_compltd_img = np.array(final_img, np.uint8)
compltd_arr.append(np.copy(np_compltd_img))
###
ver_idx = 0
for hv in range(hori_ver):
for vv in range(vert_ver):
#np_tmp_img = compltd_arr[ver_idx]
for i in range(256):
for j in range(256):
np_resized_img[vv + vert_ver*j, hv + hori_ver*i, 0] = compltd_arr[ver_idx][i, j, 0]
np_resized_img[vv + vert_ver*j, hv + hori_ver*i, 1] = compltd_arr[ver_idx][i, j, 1]
np_resized_img[vv + vert_ver*j, hv + hori_ver*i, 2] = compltd_arr[ver_idx][i, j, 2]
ver_idx += 1
###
new_compltd_img = Image.fromarray(np_resized_img)
new_compltd_img = new_compltd_img.resize((ow, oh), Image.BICUBIC)
new_compltd_img = new_compltd_img.resize((int(ow*0.5), int(oh*0.5)), Image.BICUBIC)
new_compltd_img = new_compltd_img.resize((ow, oh), Image.BICUBIC)
np_new_compltd_img = np.array(new_compltd_img)
np_oimg = np.array(oimg)
np_omsk = np.array(omsk)
np_new_compltd_img[:, :, 0] = np_new_compltd_img[:, :, 0] * (np_omsk / 255.0) + ((255.0 - np_omsk) / 255.0) * np_oimg[:, :, 0]
np_new_compltd_img[:, :, 1] = np_new_compltd_img[:, :, 1] * (np_omsk / 255.0) + ((255.0 - np_omsk) / 255.0) * np_oimg[:, :, 1]
np_new_compltd_img[:, :, 2] = np_new_compltd_img[:, :, 2] * (np_omsk / 255.0) + ((255.0 - np_omsk) / 255.0) * np_oimg[:, :, 2]
newfilename = filename.replace("_with_holes", "")
compltd_path = os.path.join(rlt_dir, newfilename)
util.save_image(np_new_compltd_img, compltd_path)
print(compltd_path)
end_time = time.time() - start_time
print('Avg Time Taken: %.3f sec' % (end_time / dataset_size))
print('done')