-
Notifications
You must be signed in to change notification settings - Fork 31
/
main_ddpir_inpainting.py
393 lines (334 loc) · 19.9 KB
/
main_ddpir_inpainting.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
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
import os.path
import cv2
import logging
import numpy as np
import torch
from datetime import datetime
from collections import OrderedDict
from utils import utils_model
from utils import utils_logger
from utils import utils_image as util
from utils.utils_inpaint import mask_generator
# from guided_diffusion import dist_util
from guided_diffusion.script_util import (
NUM_CLASSES,
model_and_diffusion_defaults,
create_model_and_diffusion,
args_to_dict,
)
def main():
# ----------------------------------------
# Preparation
# ----------------------------------------
noise_level_img = 0/255.0 # set AWGN noise level for LR image, default: 0
noise_level_model = noise_level_img # set noise level of model, default: 0
model_name = 'diffusion_ffhq_10m' # 256x256_diffusion_uncond, diffusion_ffhq_10m; set diffusino model
testset_name = 'demo_test' # set testing set, 'imagenet_val' | 'ffhq_val'
num_train_timesteps = 1000
iter_num = 20 # set number of iterations
iter_num_U = 1 # set number of inner iterations, default: 1
skip = num_train_timesteps//iter_num # skip interval
mask_name = 'gt_keep_masks/face/000000.png' # mask path for loading mask img
load_mask = False
mask_type = 'random' #['box', 'random', 'both', 'extreme']
mask_len_range = [128, 129]
mask_prob_range = [0.5, 0.5]
show_img = False # default: False
save_L = False # save LR image
save_E = True # save estimated image
save_LEH = False # save zoomed LR, E and H images
save_progressive = False # save generation process
save_progressive_mask = False # save generation process
sigma = max(0.001,noise_level_img) # noise level associated with condition y
lambda_ = 1. # key parameter lambda
sub_1_analytic = True # use analytical solution
eta = 0.0 # eta for ddim samplingn
zeta = 1.0
guidance_scale = 1.0
model_out_type = 'pred_xstart' # model output type: pred_x_prev; pred_xstart; epsilon; score
generate_mode = 'DiffPIR' # repaint; vanilla; DiffPIR
skip_type = 'quad' # uniform, quad
ddim_sample = False # sampling method
log_process = False
task_current = 'ip' # 'ip' for inpainting
n_channels = 3 # fixed
cwd = ''
model_zoo = os.path.join(cwd, 'model_zoo') # fixed
testsets = os.path.join(cwd, 'testsets') # fixed
results = os.path.join(cwd, 'results') # fixed
result_name = f'{testset_name}_{task_current}_{generate_mode}_{mask_type}_{model_name}_sigma{noise_level_img}_NFE{iter_num}_eta{eta}_zeta{zeta}_lambda{lambda_}'
model_path = os.path.join(model_zoo, model_name+'.pt')
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
torch.cuda.empty_cache()
calc_LPIPS = True
# noise schedule
beta_start = 0.1 / 1000
beta_end = 20 / 1000
betas = np.linspace(beta_start, beta_end, num_train_timesteps, dtype=np.float32)
betas = torch.from_numpy(betas).to(device)
alphas = 1.0 - betas
alphas_cumprod = np.cumprod(alphas.cpu(), axis=0)
sqrt_alphas_cumprod = torch.sqrt(alphas_cumprod)
sqrt_1m_alphas_cumprod = torch.sqrt(1. - alphas_cumprod)
reduced_alpha_cumprod = torch.div(sqrt_1m_alphas_cumprod, sqrt_alphas_cumprod) # equivalent noise sigma on image
noise_model_t = utils_model.find_nearest(reduced_alpha_cumprod, 2 * noise_level_model)
noise_model_t = 0
noise_inti_img = 50 / 255
t_start = utils_model.find_nearest(reduced_alpha_cumprod, 2 * noise_inti_img) # start timestep of the diffusion process
t_start = num_train_timesteps - 1
# ----------------------------------------
# L_path, E_path, H_path, mask_path
# ----------------------------------------
L_path = os.path.join(testsets, testset_name) # L_path, for Low-quality images
E_path = os.path.join(results, result_name) # E_path, for Estimated images
mask_path = os.path.join(testsets, mask_name) # mask_path, for mask images
util.mkdir(E_path)
logger_name = result_name
utils_logger.logger_info(logger_name, log_path=os.path.join(E_path, logger_name+'.log'))
logger = logging.getLogger(logger_name)
# ----------------------------------------
# load model
# ----------------------------------------
model_config = dict(
model_path=model_path,
num_channels=128,
num_res_blocks=1,
attention_resolutions="16",
) if model_name == 'diffusion_ffhq_10m' \
else dict(
model_path=model_path,
num_channels=256,
num_res_blocks=2,
attention_resolutions="8,16,32",
)
args = utils_model.create_argparser(model_config).parse_args([])
model, diffusion = create_model_and_diffusion(
**args_to_dict(args, model_and_diffusion_defaults().keys()))
# model.load_state_dict(
# dist_util.load_state_dict(args.model_path, map_location="cpu")
# )
model.load_state_dict(torch.load(args.model_path, map_location="cpu"))
model.eval()
for k, v in model.named_parameters():
v.requires_grad = False
model = model.to(device)
logger.info('model_name:{}, mask_type:{}, image sigma:{:.3f}, model sigma:{:.3f}'.format(model_name, mask_type, noise_level_img, noise_level_model))
logger.info('eta:{:.3f}, zeta:{:.3f}, lambda:{:.3f}, guidance_scale:{:.2f} '.format(eta, zeta, lambda_, guidance_scale))
logger.info('start step:{}, skip_type:{}, skip interval:{}, skipstep analytic steps:{}'.format(t_start, skip_type, skip, noise_model_t))
logger.info('Model path: {:s}'.format(model_path))
logger.info(L_path)
L_paths = util.get_image_paths(L_path)
if calc_LPIPS:
import lpips
loss_fn_vgg = lpips.LPIPS(net='vgg').to(device)
def test_rho(lambda_=lambda_,model_out_type_=model_out_type,zeta=zeta):
model_out_type = model_out_type_
logger.info('eta:{:.3f}, zeta:{:.3f}, lambda:{:.3f}, guidance_scale:{:.2f}'.format(eta, zeta, lambda_, guidance_scale))
test_results = OrderedDict()
test_results['psnr'] = []
if calc_LPIPS:
test_results['lpips'] = []
for idx, img in enumerate(L_paths):
# --------------------------------
# (1) get img_H and img_L
# --------------------------------
idx += 1
img_name, ext = os.path.splitext(os.path.basename(img))
img_H = util.imread_uint(img, n_channels=n_channels)
# --------------------------------
# (2) initialize x
# --------------------------------
if load_mask:
mask = util.imread_uint(mask_path, n_channels=n_channels).astype(bool)
else:
mask_gen = mask_generator(mask_type=mask_type, mask_len_range=mask_len_range, mask_prob_range=mask_prob_range)
np.random.seed(seed=0) # for reproducibility
mask = mask_gen(util.uint2tensor4(img_H)).numpy()
mask = np.squeeze(mask)
mask = np.transpose(mask, (1, 2, 0))
img_L = img_H * mask / 255. #(256,256,3) [0,1]
np.random.seed(seed=0) # for reproducibility
img_L = img_L * 2 - 1
img_L += np.random.normal(0, noise_level_img * 2, img_L.shape) # add AWGN
img_L = img_L / 2 + 0.5
img_L = img_L * mask
y = util.single2tensor4(img_L).to(device) #(1,3,256,256)
y = y * 2 -1 # [-1,1]
mask = util.single2tensor4(mask.astype(np.float32)).to(device)
# for y with given noise level, add noise from t_y
t_y = utils_model.find_nearest(reduced_alpha_cumprod, 2 * noise_level_img)
sqrt_alpha_effective = sqrt_alphas_cumprod[t_start] / sqrt_alphas_cumprod[t_y]
x = sqrt_alpha_effective * y + torch.sqrt(sqrt_1m_alphas_cumprod[t_start]**2 - \
sqrt_alpha_effective**2 * sqrt_1m_alphas_cumprod[t_y]**2) * torch.randn_like(y)
# x = sqrt_alphas_cumprod[t_start] * y + sqrt_1m_alphas_cumprod[t_start] * torch.randn_like(y)
# --------------------------------
# (3) get rhos and sigmas
# --------------------------------
sigmas = []
sigma_ks = []
rhos = []
for i in range(num_train_timesteps):
sigmas.append(reduced_alpha_cumprod[num_train_timesteps-1-i])
if model_out_type == 'pred_xstart':
sigma_ks.append((sqrt_1m_alphas_cumprod[i]/sqrt_alphas_cumprod[i]))
elif model_out_type == 'pred_x_prev':
sigma_ks.append(torch.sqrt(betas[i]/alphas[i]))
rhos.append(lambda_*(sigma**2)/(sigma_ks[i]**2))
rhos, sigmas, sigma_ks = torch.tensor(rhos).to(device), torch.tensor(sigmas).to(device), torch.tensor(sigma_ks).to(device)
# --------------------------------
# (4) main iterations
# --------------------------------
progress_img = []
# create sequence of timestep for sampling
if skip_type == 'uniform':
seq = [i*skip for i in range(iter_num)]
if skip > 1:
seq.append(num_train_timesteps-1)
elif skip_type == "quad":
seq = np.sqrt(np.linspace(0, num_train_timesteps**2, iter_num))
seq = [int(s) for s in list(seq)]
seq[-1] = seq[-1] - 1
progress_seq = seq[::(len(seq)//10)]
progress_seq.append(seq[-1])
# reverse diffusion for one image from random noise
for i in range(len(seq)):
curr_sigma = sigmas[seq[i]].cpu().numpy()
# time step associated with the noise level sigmas[i]
t_i = utils_model.find_nearest(reduced_alpha_cumprod,curr_sigma)
# skip iters
if t_i > t_start:
continue
for u in range(iter_num_U):
# --------------------------------
# step 1, reverse diffsuion step
# --------------------------------
# add noise, make the image noise level consistent in pixel level
if generate_mode == 'repaint':
x = (sqrt_alphas_cumprod[t_i] * y + sqrt_1m_alphas_cumprod[t_i] * torch.randn_like(x)) * mask \
+ (1-mask) * x
# solve equation 6b with one reverse diffusion step
if model_out_type == 'pred_xstart':
x0 = utils_model.model_fn(x, noise_level=curr_sigma*255, model_out_type=model_out_type, \
model_diffusion=model, diffusion=diffusion, ddim_sample=ddim_sample, alphas_cumprod=alphas_cumprod)
else:
x = utils_model.model_fn(x, noise_level=curr_sigma*255, model_out_type=model_out_type, \
model_diffusion=model, diffusion=diffusion, ddim_sample=ddim_sample, alphas_cumprod=alphas_cumprod)
# x = utils_model.test_mode(model_fn, x, mode=0, refield=32, min_size=256, modulo=16, noise_level=sigmas[i].cpu().numpy()*255)
# --------------------------------
# step 2, closed-form solution
# --------------------------------
# analytic solution
if (generate_mode == 'DiffPIR') and not (seq[i] == seq[-1]):
# solve sub-problem
if sub_1_analytic:
if model_out_type == 'pred_xstart':
# when noise level less than given image noise, skip
if i < num_train_timesteps-noise_model_t:
x0_p = (mask*y + rhos[t_i].float()*x0).div(mask+rhos[t_i])
x0 = x0 + guidance_scale * (x0_p-x0)
else:
model_out_type = 'pred_x_prev'
x0 = utils_model.model_fn(x, noise_level=curr_sigma*255, model_out_type=model_out_type, \
model_diffusion=model, diffusion=diffusion, ddim_sample=ddim_sample, alphas_cumprod=alphas_cumprod)
pass
elif model_out_type == 'pred_x_prev':
# when noise level less than given image noise, skip
if i < num_train_timesteps-noise_model_t:
x = (mask*y + rhos[t_i].float()*x).div(mask+rhos[t_i]) # y-->yt ?
else:
pass
else:
# TODO: first order solver
# x = x - 1 / (2*rhos[t_i]) * (x - y_t) * mask
pass
if (model_out_type == 'pred_xstart') and not (seq[i] == seq[-1]):
# x = sqrt_alphas_cumprod[t_i] * (x) + (sqrt_1m_alphas_cumprod[t_i]) * torch.randn_like(x) # x = sqrt_alphas_cumprod[t_i] * (x) + (sqrt_1m_alphas_cumprod[t_i]) * torch.randn_like(x)
t_im1 = utils_model.find_nearest(reduced_alpha_cumprod,sigmas[seq[i+1]].cpu().numpy())
# calculate \hat{\eposilon}
eps = (x - sqrt_alphas_cumprod[t_i] * x0) / sqrt_1m_alphas_cumprod[t_i]
eta_sigma = eta * sqrt_1m_alphas_cumprod[t_im1] / sqrt_1m_alphas_cumprod[t_i] * torch.sqrt(betas[t_i])
x = sqrt_alphas_cumprod[t_im1] * x0 + np.sqrt(1-zeta) * (torch.sqrt(sqrt_1m_alphas_cumprod[t_im1]**2 - eta_sigma**2) * eps \
+ eta_sigma * torch.randn_like(x)) + np.sqrt(zeta) * sqrt_1m_alphas_cumprod[t_im1] * torch.randn_like(x)
# set back to x_t from x_{t-1}
if u < iter_num_U-1 and seq[i] != seq[-1]:
# x = torch.sqrt(alphas[t_i]) * x + torch.sqrt(betas[t_i]) * torch.randn_like(x)
sqrt_alpha_effective = sqrt_alphas_cumprod[t_i] / sqrt_alphas_cumprod[t_im1]
x = sqrt_alpha_effective * x + torch.sqrt(sqrt_1m_alphas_cumprod[t_i]**2 - \
sqrt_alpha_effective**2 * sqrt_1m_alphas_cumprod[t_im1]**2) * torch.randn_like(x)
# save the process
x_0 = (x/2+0.5)
if save_progressive and (seq[i] in progress_seq):
x_show = x_0.clone().detach().cpu().numpy() #[0,1]
x_show = np.squeeze(x_show)
if x_show.ndim == 3:
x_show = np.transpose(x_show, (1, 2, 0))
progress_img.append(x_show)
if log_process:
logger.info('{:>4d}, steps: {:>4d}, np.max(x_show): {:.4f}, np.min(x_show): {:.4f}'.format(seq[i], t_i, np.max(x_show), np.min(x_show)))
if show_img:
util.imshow(x_show)
# recover conditional part
if generate_mode in ['repaint','DiffPIR']:
x[mask.to(torch.bool)] = y[mask.to(torch.bool)]
# --------------------------------
# (4) save process
# --------------------------------
img_E = util.tensor2uint(x_0)
psnr = util.calculate_psnr(img_E, img_H, border=0) # change with your own border
test_results['psnr'].append(psnr)
if calc_LPIPS:
img_H_tensor = np.transpose(img_H, (2, 0, 1))
img_H_tensor = torch.from_numpy(img_H_tensor)[None,:,:,:].to(device)
img_H_tensor = img_H_tensor / 255 * 2 -1
lpips_score = loss_fn_vgg(x_0.detach()*2-1, img_H_tensor)
lpips_score = lpips_score.cpu().detach().numpy()[0][0][0][0]
test_results['lpips'].append(lpips_score)
logger.info('{:->4d}--> {:>10s} PSNR: {:.4f}dB LPIPS: {:.4f} ave LPIPS: {:.4f}'.format(idx, img_name+ext, psnr, lpips_score, sum(test_results['lpips']) / len(test_results['lpips'])))
else:
logger.info('{:->4d}--> {:>10s} PSNR: {:.4f}dB'.format(idx, img_name+ext, psnr))
pass
if save_E:
util.imsave(img_E, os.path.join(E_path, img_name+'_'+model_name+ext))
if save_L:
util.imsave(util.single2uint(img_L), os.path.join(E_path, img_name+'_L'+ext))
if save_LEH:
util.imsave(np.concatenate([util.single2uint(img_L), img_E, img_H], axis=1), os.path.join(E_path, img_name+model_name+'_LEH'+ext))
if save_progressive:
now = datetime.now()
current_time = now.strftime("%Y_%m_%d_%H_%M_%S")
if generate_mode in ['repaint','DiffPIR']:
mask = np.squeeze(mask.cpu().numpy())
if mask.ndim == 3:
mask = np.transpose(mask, (1, 2, 0))
img_total = cv2.hconcat(progress_img)
if show_img:
util.imshow(img_total,figsize=(80,4))
util.imsave(img_total*255., os.path.join(E_path, img_name+'_process_lambda_{:.3f}_{}{}'.format(lambda_,current_time,ext)))
images = []
y_t = np.squeeze((y/2+0.5).cpu().numpy())
if y_t.ndim == 3:
y_t = np.transpose(y_t, (1, 2, 0))
if generate_mode in ['repaint','DiffPIR']:
for x in progress_img:
images.append((y_t)* mask+ (1-mask) * x)
img_total = cv2.hconcat(images)
if show_img:
util.imshow(img_total,figsize=(80,4))
if save_progressive_mask:
util.imsave(img_total*255., os.path.join(E_path, img_name+'_process_mask_lambda_{:.3f}_{}{}'.format(lambda_,current_time,ext)))
# --------------------------------
# Average PSNR and LPIPS
# --------------------------------
ave_psnr = sum(test_results['psnr']) / len(test_results['psnr'])
logger.info('------> Average PSNR of ({}), sigma: ({:.3f}): {:.4f} dB'.format(testset_name, noise_level_model, ave_psnr))
if calc_LPIPS:
ave_lpips = sum(test_results['lpips']) / len(test_results['lpips'])
logger.info('------> Average LPIPS of ({}), sigma: ({:.3f}): {:.4f}'.format(testset_name, noise_level_model, ave_lpips))
# experiments
lambdas = [lambda_*i for i in range(1,2)]
for lambda_ in lambdas:
#for zeta_i in [0,0.3,0.8,0.9,1.0]:
for zeta_i in [zeta*i for i in range(1,2)]:
test_rho(lambda_, zeta=zeta_i)
if __name__ == '__main__':
main()