-
Notifications
You must be signed in to change notification settings - Fork 0
/
ddp_engine_test.py
217 lines (195 loc) · 10.7 KB
/
ddp_engine_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
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
import torch
from torchvision.transforms import functional as TF
import os
import cv2
import glob
import numpy as np
from tqdm import tqdm
from collections import OrderedDict
from utils import acc_utils
from dataset import make_some_noise
@torch.no_grad()
def test_one_epoch(rank, gpu, model, epoch, args, opts):
model.eval()
if args.task == 'lightweight_sr':
default_dataset_list = ['Set5', 'Set14', 'BSDS100', 'urban100', 'manga109']
if args.world_size==4:
dataset_list = [default_dataset_list[rank+1]]
if rank==0:
dataset_list.insert(0, default_dataset_list[0])
elif args.world_size==2:
if rank==0:
dataset_list = default_dataset_list[:4]
else:
dataset_list = default_dataset_list[4:]
elif args.target_mode == 'light_dn':
default_dataset_list = ['CBSD68', 'Kodak24', 'McMaster', 'urban100']
if args.world_size==4:
dataset_list = [default_dataset_list[rank]]
elif args.world_size==2:
if rank==0:
dataset_list = default_dataset_list[:3]
else:
dataset_list = default_dataset_list[3:]
elif args.target_mode == 'light_graydn':
default_dataset_list = ['Set12', 'CBSD68', 'urban100']
if args.world_size==4:
if rank != 0:
dataset_list = [default_dataset_list[rank-1]]
else:
dataset_list = ['Set12']
elif args.world_size==2:
if rank==0:
dataset_list = default_dataset_list[:2]
else:
dataset_list = default_dataset_list[2:]
elif args.target_mode == 'light_lle':
default_dataset_list = ['LOL', 'VELOL-cap']
if args.world_size==4:
if rank < 2:
dataset_list = default_dataset_list[:1]
else:
dataset_list = default_dataset_list[1:]
elif args.world_size==2:
if rank==0:
dataset_list = default_dataset_list[:1]
else:
dataset_list = default_dataset_list[1:]
elif args.target_mode == 'light_dr':
default_dataset_list = ['Rain100H', 'Test100']
if args.world_size==4:
if rank < 2:
dataset_list = default_dataset_list[:1]
else:
dataset_list = default_dataset_list[1:]
elif args.world_size==2:
if rank==0:
dataset_list = default_dataset_list[:1]
else:
dataset_list = default_dataset_list[1:]
if 'RAMiT' in args.model_name:
min_multiple = (4*opts['window_size'], 4*opts['window_size'])
dataset_list = default_dataset_list if args.world_size==1 else dataset_list
for dd, dataset in enumerate(dataset_list):
test_results = OrderedDict()
if args.task=='lightweight_sr':
folder_lq = f'../testsets/{dataset}/LR_bicubic/X{args.scale}/'
folder_hq = f'../testsets/{dataset}/HR/'
test_degrade = ('',)
save_dir = f'results/{args.model_name}_{args.task}_x{args.scale}/{args.model_time}'
test_results['psnr_y'] = []
test_results['ssim_y'] = []
psnr_y, ssim_y = 0, 0
elif args.task == 'lightweight_dn':
folder_hq = f'../testsets/{dataset}/HQ/'
test_degrade = (15,25,50)
save_dir = f'results/{args.model_name}_{args.task}/{args.model_time}'
test_results['psnr'] = []
test_results['ssim'] = []
psnr, ssim = 0, 0
elif args.task == 'lightweight_lle':
folder_lq = f'../testsets/{dataset}/LQ/'
folder_hq = f'../testsets/{dataset}/HQ/'
test_degrade = ('',)
save_dir = f'results/{args.model_name}_{args.task}/{args.model_time}'
test_results['psnr'] = []
test_results['ssim'] = []
psnr, ssim = 0, 0
elif args.task == 'lightweight_dr':
folder_lq = f'../testsets/{dataset}/LQ/'
folder_hq = f'../testsets/{dataset}/HQ/'
test_degrade = ('',)
save_dir = f'results/{args.model_name}_{args.task}/{args.model_time}'
test_results['psnr_y'] = []
test_results['ssim_y'] = []
psnr_y, ssim_y = 0, 0
border = args.scale if args.task == 'lightweight_sr' else 0
# setup result dictionary
os.makedirs(save_dir, exist_ok=True)
path_list = sorted(glob.glob(os.path.join(folder_hq, '*.npy')))
imgname_maxlen = max([len(os.path.splitext(os.path.basename(p))[0]) for p in path_list])
for degrade in test_degrade:
for data_iter, path in enumerate(tqdm(path_list)):
# read image
if 'gray' not in args.target_mode:
img_hq = torch.from_numpy(np.load(path))/255
else:
img_hq = np.transpose(np.load(path), (1,2,0))
img_hq = cv2.cvtColor(img_hq, cv2.COLOR_RGB2GRAY)
img_hq = torch.from_numpy(img_hq).unsqueeze(0)/255
imgname, imgext = os.path.splitext(os.path.basename(path))
if args.task=='lightweight_sr':
path_lq = os.path.join(folder_lq, f'{imgname}x{args.scale}{imgext}')
img_lq = torch.from_numpy(np.load(path_lq)).unsqueeze(0)/255
elif args.task=='lightweight_dn':
img_lq = torch.clone(img_hq)
img_lq = make_some_noise(img_lq, (degrade,), seed=0).unsqueeze(0)
elif args.task in ['lightweight_lle', 'lightweight_dr']:
path_lq = os.path.join(folder_lq, f'{imgname}{imgext}')
img_lq = torch.from_numpy(np.load(path_lq)).unsqueeze(0)/255
# inference
# pad input image to be a multiple of window_size X final_patch_size
_, _, lqh, lqw = img_lq.size()
padw = min_multiple[1] - (lqw%min_multiple[1]) if lqw%min_multiple[1]!=0 else 0
padh = min_multiple[0] - (lqh%min_multiple[0]) if lqh%min_multiple[0]!=0 else 0
img_lq = TF.pad(img_lq, (0,0,padw,padh), padding_mode='symmetric')
img_rc = model(img_lq.to(gpu))
img_rc = img_rc[..., :lqh*args.scale, :lqw*args.scale]
# save image
img_rc = img_rc[0].detach().cpu().clamp(0,1).numpy()
if 'gray' not in args.target_mode:
img_rc = np.transpose(img_rc[[2, 1, 0],:,:], (1, 2, 0)) if img_rc.ndim == 3 else img_rc # CHW-RGB to HWC-BGR
else:
img_rc = np.transpose(img_rc, (1, 2, 0)) if img_rc.ndim == 3 else img_rc # CHW-RGB to HWC-BGR
img_rc = (img_rc * 255.0).round().astype(np.uint8) # float32 to uint8
if args.result_image_save:
if args.task=='lightweight_sr':
cv2.imwrite(f'{save_dir}/{dataset}_{imgname}_x{args.scale}_{args.model_name}.png', img_rc)
elif args.task=='lightweight_dn':
cv2.imwrite(f'{save_dir}/{dataset}_{imgname}_sigma{degrade}_{args.model_name}.png', img_rc)
elif args.task in ['lightweight_lle', 'lightweight_dr']:
cv2.imwrite(f'{save_dir}/{dataset}_{imgname}_{args.model_name}.png', img_rc)
# evaluate psnr/ssim
if 'gray' not in args.target_mode:
img_hq = img_hq.permute(1,2,0)[:,:,[2,1,0]].numpy() # CHW-RGB to HWC-BGR
else:
img_hq = img_hq.permute(1,2,0).numpy() # CHW-Gr to HWC-Gr (Gr: Gray)
img_hq = (img_hq * 255.0).round().astype(np.uint8) # float32 to uint8
img_hq = img_hq[:lqh*args.scale,:lqw*args.scale,:] # crop HQ
img_hq = np.squeeze(img_hq) if 'gray' not in args.target_mode else img_hq
if args.task in ['lightweight_sr', 'lightweight_dr']:
psnr_y = acc_utils.calculate_psnr(img_rc, img_hq, crop_border=border, test_y_channel=True)
ssim_y = acc_utils.calculate_ssim(img_rc, img_hq, crop_border=border, test_y_channel=True)
test_results['psnr_y'].append(psnr_y)
test_results['ssim_y'].append(ssim_y)
elif args.task in ['lightweight_dn', 'lightweight_lle']:
psnr = acc_utils.calculate_psnr(img_rc, img_hq, crop_border=border, test_y_channel=False)
ssim = acc_utils.calculate_ssim(img_rc, img_hq, crop_border=border, test_y_channel=False)
test_results['psnr'].append(psnr)
test_results['ssim'].append(ssim)
deg = f'_{degrade}' if degrade!='' else ''
with open(f'./logs/{args.model_time}_test_{dataset}{deg}.txt', 'a') as f:
if data_iter==0: f.writelines(f'[[{epoch+1}]]\n')
for _ in range(imgname_maxlen-len(imgname)): imgname+=' '
f.writelines(f'{dataset:10s} {data_iter+1:3d} {imgname} - ')
if args.task in ['lightweight_sr', 'lightweight_dr']:
f.writelines(f'PSNR_Y: {psnr_y:.2f}, SSIM_Y: {ssim_y:.4f}\n')
elif args.task in ['lightweight_dn', 'lightweight_lle']:
f.writelines(f'PSNR: {psnr:.2f}, SSIM: {ssim:.4f}\n')
if data_iter+1 == len(path_list): f.writelines('\n')
# summarize psnr/ssim
with open(f'./logs/{args.model_time}_test{deg}.txt', 'a') as f:
if (rank==0 and dd==0) or (rank==1 and dd==0 and args.target_mode in ['light_graydn', 'light_lle', 'light_dr'] and args.world_size==4):
f.writelines(f'[[{epoch+1}]]\n')
if args.task in ['lightweight_sr', 'lightweight_dr']:
avg_psnr_y = sum(test_results['psnr_y']) / len(test_results['psnr_y'])
avg_ssim_y = sum(test_results['ssim_y']) / len(test_results['ssim_y'])
f.writelines(f'{dataset} {args.target_mode} - PSNR_Y/SSIM_Y: {avg_psnr_y:.2f}/{avg_ssim_y:.4f}\n')
elif args.task in ['lightweight_dn', 'lightweight_lle']:
avg_psnr = sum(test_results['psnr']) / len(test_results['psnr'])
avg_ssim = sum(test_results['ssim']) / len(test_results['ssim'])
f.writelines(f'{dataset} {args.target_mode} - PSNR/SSIM: {avg_psnr:.2f}/{avg_ssim:.4f}\n')
if rank==args.world_size-1 and dd+1==len(dataset_list):
f.writelines('\n')
for k in test_results.keys():
test_results[k] = []