-
Notifications
You must be signed in to change notification settings - Fork 0
/
metrics_python.py
111 lines (86 loc) · 5.04 KB
/
metrics_python.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
import cv2
import glob
import numpy as np
import os.path as osp
import torch
import lpips
from torchvision.transforms.functional import normalize
from scipy.io import savemat
import os
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
def img2tensor(img):
img = img.transpose(2, 0, 1).astype('float32')/255
tensor = torch.from_numpy(img)
mean = [0.5, 0.5, 0.5]
std = [0.5, 0.5, 0.5]
tensor_norm = normalize(tensor, mean, std).cuda()
return tensor_norm
def main():
data_name = 'AID'
data_train = 'bicubic'
data_test = 'bicubic'
folder_hr = 'D:/Datasets/AID_all/generated/clean/valid_real_tdsr/HR'
folders_sr = [ 'D:/#11/codes/Real-SR-master/codes/results_'+data_name+'/'+data_test+'_upsample'+'/'+data_name,
'D:/#11/codes/Real-SR-master/codes/results_'+data_name+'/srcnn_'+data_train+'_'+data_test+'/'+data_name,
'D:/#11/codes/Real-SR-master/codes/results_'+data_name+'/vdsr_'+data_train+'_'+data_test+'/'+data_name,
# 'D:/#11/codes/Real-SR-master/codes/results_'+data_name+'/lapsrn_'+data_train+'_'+data_test+'/'+data_name,
'D:/#11/codes/Real-SR-master/codes/results_'+data_name+'/ddbpn_'+data_train+'_'+data_test+'/'+data_name,
'D:/#11/codes/Real-SR-master/codes/results_'+data_name+'/edsr_'+data_train+'_'+data_test+'/'+data_name,
'D:/#11/codes/Real-SR-master/codes/results_'+data_name+'/srgan_'+data_train+'_'+data_test+'/'+data_name,
'D:/#11/codes/Real-SR-master/codes/results_'+data_name+'/drln_'+data_train+'_'+data_test+'/'+data_name,
'D:/#11/codes/Real-SR-master/codes/results_'+data_name+'/rban_unet_'+data_train+'_'+data_test+'/'+data_name]
loss_fn_alex = lpips.LPIPS(net='alex').cuda() # best forward scores
img_hr_list = sorted(glob.glob(osp.join(folder_hr, '*')))
num_methods = len(folders_sr)
num_imgs = len(img_hr_list)
lpips_all = np.zeros((num_methods, num_imgs))
for idx_img in range(num_imgs):
# for idx_img in range(10):
if ( (idx_img+1) % 10 == 0):
print(idx_img+1)
path_hr = img_hr_list[idx_img]
img_hr = cv2.cvtColor(cv2.imread(path_hr, cv2.IMREAD_UNCHANGED) ,cv2.COLOR_BGR2RGB)
tensor_norm_hr = img2tensor(img_hr)
file_name = osp.basename(path_hr)
for idx_method in range(num_methods):
path_sr = osp.join(folders_sr[idx_method], file_name)
img_sr = cv2.cvtColor(cv2.imread(path_sr, cv2.IMREAD_UNCHANGED) ,cv2.COLOR_BGR2RGB)
tensor_norm_sr = img2tensor(img_sr)
lpips_all[idx_method, idx_img] = loss_fn_alex(tensor_norm_hr, tensor_norm_sr).detach().cpu().numpy()
lpips_mean = np.mean(lpips_all, axis=1)
mat_name = 'D:/#11/codes/Real-SR-master/codes/matlab_metrics/lpips_'+data_name+'_'+data_train+'_'+data_test+'.mat'
savemat(mat_name, {'lpips':lpips_mean})
def main_ablation():
folder_hr = 'D:/Datasets/AID_all/generated/clean/valid_real_tdsr/HR'
folders_sr = [ 'D:/#11/codes/Real-SR-master/codes/results_ablation/real_upsample/AID/',
'D:/#11/codes/Real-SR-master/codes/results_ablation/rban_unet_bicubic_real/AID/',
'D:/#11/codes/Real-SR-master/codes/results_ablation/rban_unet_noblur/AID/',
'D:/#11/codes/Real-SR-master/codes/results_ablation/rban_unet_nonoise/AID/',
'D:/#11/codes/Real-SR-master/codes/results_ablation/rgn_real_real/AID/',
'D:/#11/codes/Real-SR-master/codes/results_ablation/rban_real_real/AID/',
'D:/#11/codes/Real-SR-master/codes/results_ablation/rban_vgg_real_real/AID/',
'D:/#11/codes/Real-SR-master/codes/results_ablation/rban_unet_real_real/AID/']
loss_fn_alex = lpips.LPIPS(net='alex').cuda() # best forward scores
img_hr_list = sorted(glob.glob(osp.join(folder_hr, '*')))
num_methods = len(folders_sr)
num_imgs = len(img_hr_list)
lpips_all = np.zeros((num_methods, num_imgs))
for idx_img in range(num_imgs):
# for idx_img in range(10):
if ( (idx_img+1) % 10 == 0):
print(idx_img+1)
path_hr = img_hr_list[idx_img]
img_hr = cv2.cvtColor(cv2.imread(path_hr, cv2.IMREAD_UNCHANGED) ,cv2.COLOR_BGR2RGB)
tensor_norm_hr = img2tensor(img_hr)
file_name = osp.basename(path_hr)
for idx_method in range(num_methods):
path_sr = osp.join(folders_sr[idx_method], file_name)
img_sr = cv2.cvtColor(cv2.imread(path_sr, cv2.IMREAD_UNCHANGED) ,cv2.COLOR_BGR2RGB)
tensor_norm_sr = img2tensor(img_sr)
lpips_all[idx_method, idx_img] = loss_fn_alex(tensor_norm_hr, tensor_norm_sr).detach().cpu().numpy()
lpips_mean = np.mean(lpips_all, axis=1)
mat_name = 'D:/#11/codes/Real-SR-master/codes/matlab_ablation_metrics/lpips_ablation.mat'
savemat(mat_name, {'lpips':lpips_mean})
if __name__ == '__main__':
main()
# main_ablation()