-
Notifications
You must be signed in to change notification settings - Fork 1
/
test.py
164 lines (135 loc) · 6.43 KB
/
test.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
import torch
import os
import time
import numpy as np
import pandas as pd
from skimage.metrics import peak_signal_noise_ratio, structural_similarity
import cv2
import skimage.color as color
import utils
from model import HVLF
def test_main(image_path, model_path, view_n_ori, view_n, scale, cut_margin=15, gpu_no=0, is_fine_tune=False,
save_img=False):
os.environ["CUDA_VISIBLE_DEVICES"] = str(gpu_no)
torch.backends.cudnn.benchmark = True
print('=' * 40)
print('build network...')
model = HVLF(scale=scale)
utils.get_parameter_number(model)
# return
model.cuda()
model.eval()
print('done')
print('=' * 40)
print('load model...')
if is_fine_tune:
state_dict = torch.load(model_path + 'HVLF_' + str(scale) + '_' + str(view_n) + '_ft.pkl')
else:
state_dict = torch.load(model_path + 'HVLF_' + str(scale) + '_' + str(view_n) + '.pkl')
model.load_state_dict(state_dict)
print('done')
print('=' * 40)
print('create save directory...')
if is_fine_tune:
save_path = '../result/scale{}view{}_fine_tune_result/'.format(scale, view_n)
else:
save_path = '../result/scale{}view{}_result/'.format(scale, view_n)
if not os.path.exists(save_path):
os.makedirs(save_path)
print('path: ', save_path)
print('done')
print('=' * 40)
print('predict image...')
xls_list = []
psnr_list = []
ssim_list = []
time_list = []
image_list = os.listdir(image_path)
image_list.sort()
for index, image_name in enumerate(image_list):
print('-' * 100)
print('[{}/{}]'.format(index + 1, len(image_list)), image_name)
gt_hr_y, lr_ycbcr = utils.image_input(image_path + image_name, scale, view_n_ori, view_n)
lr_y, gt_hr_y, lr_cbcr = lr_ycbcr[:, :, :, :, :, 0] / 255.0, gt_hr_y / 255.0, lr_ycbcr[:, :, :, :, :, 1:3]
hr_y, time_ = predict_y(lr_y, model)
time_list.append(time_)
hr_y = hr_y.cpu().numpy()
hr_y = np.clip(hr_y, 16.0 / 255.0, 235.0 / 255.0)
gt_hr_y = np.clip(gt_hr_y, 16.0 / 255.0, 235.0 / 255.0)
psnr_view_list = []
ssim_view_list = []
for i in range(view_n):
for j in range(view_n):
if cut_margin == 0:
psnr_view = peak_signal_noise_ratio(hr_y[0, i, j, :, :],
gt_hr_y[0, i, j, :, :], data_range=1)
psnr_view_list.append(psnr_view)
ssim_view = structural_similarity(hr_y[0, i, j, :, :],
gt_hr_y[0, i, j, :, :], data_range=1)
ssim_view_list.append(ssim_view)
else:
psnr_view = peak_signal_noise_ratio(hr_y[0, i, j, cut_margin:-cut_margin, cut_margin:-cut_margin],
gt_hr_y[0, i, j, cut_margin:-cut_margin,
cut_margin:-cut_margin], data_range=1)
psnr_view_list.append(psnr_view)
ssim_view = structural_similarity(hr_y[0, i, j, cut_margin:-cut_margin, cut_margin:-cut_margin],
gt_hr_y[0, i, j, cut_margin:-cut_margin, cut_margin:-cut_margin],
data_range=1)
ssim_view_list.append(ssim_view)
print('{:6.4f}/{:6.4f}'.format(psnr_view, ssim_view), end='\t\t')
print('')
if save_img:
result_image_path = save_path + image_name[0:-4] + '/'
if not os.path.exists(result_image_path):
os.makedirs(result_image_path)
hr_cbcr = predict_cbcr(lr_cbcr, scale, view_n)
for i in range(view_n):
for j in range(view_n):
hr_y_item = np.clip(hr_y[0, i, j, :, :] * 255.0, 16.0, 235.0)
hr_y_item = hr_y_item[:, :, np.newaxis]
hr_cb_item = hr_cbcr[0, i, j, :, :, 0:1]
hr_cr_item = hr_cbcr[0, i, j, :, :, 1:2]
hr_ycbcr_item = np.concatenate((hr_y_item, hr_cb_item, hr_cr_item), 2)
hr_rgb_item = color.ycbcr2rgb(hr_ycbcr_item) * 255.0
hr_rgb_item = hr_rgb_item[:, :, ::-1]
img_save_path = result_image_path + str(i) + str(j) + '.png'
cv2.imwrite(img_save_path, hr_rgb_item)
psnr_ = np.mean(psnr_view_list)
psnr_list.append(psnr_)
ssim_ = np.mean(ssim_view_list)
ssim_list.append(ssim_)
print('PSNR: {:.4f} SSIM: {:.4f} time: {:.4f}'.format(psnr_, ssim_, time_))
xls_list.append([image_name, psnr_, ssim_, time_])
xls_list.append(['average', np.mean(psnr_list), np.mean(ssim_list), np.mean(time_list)])
xls_list = np.array(xls_list)
result = pd.DataFrame(xls_list, columns=['image', 'psnr', 'ssim', 'time'])
result.to_csv(save_path + 'result_s{}n{}_{}.csv'.format(scale, view_n, int(time.time())))
print('-' * 100)
print('Average: PSNR: {:.4f}, SSIM: {:.4f}, TIME: {:.4f}'.format(np.mean(psnr_list), np.mean(ssim_list),
np.mean(time_list)))
print('all done')
def predict_y(lr_y, model):
with torch.no_grad():
lr_y = torch.from_numpy(lr_y.copy())
lr_y = lr_y.cuda()
time_item_start = time.time()
hr_y = model(lr_y)
return hr_y, time.time() - time_item_start
def predict_cbcr(lr_cbcr, scale, view_n):
hr_cbcr = np.zeros((1, view_n, view_n, lr_cbcr.shape[3] * scale, lr_cbcr.shape[4] * scale, 2))
for i in range(view_n):
for j in range(view_n):
image_bicubic = cv2.resize(lr_cbcr[0, i, j, :, :, :],
(lr_cbcr.shape[4] * scale, lr_cbcr.shape[3] * scale),
interpolation=cv2.INTER_CUBIC)
hr_cbcr[0, i, j, :, :, :] = image_bicubic
return hr_cbcr
if __name__ == '__main__':
test_main(image_path='../../Dataset/test_hci1/', # .png
model_path='../models/',
view_n_ori=9,
view_n=3,
scale=2,
gpu_no=3,
is_fine_tune=True,
save_img=False)