-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathDenoise_enI_train.py
227 lines (185 loc) · 10.2 KB
/
Denoise_enI_train.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
import torch
import torch.nn as nn
from torch.utils.data import DataLoader
from torch.autograd import Variable
import numpy as np
import cv2
from PIL import Image
import matplotlib.pyplot as plt
import argparse, datetime, time
import os
from glob import glob
from MyDataset import *
from myutils import *
from loss import *
from network_unet import UNetRes as net
from model import DecomNet_RTV, EnhanceNet_I
import pytorch_ssim
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
def train_Denoising_enI(dataloader, eval_floder, numBatch):
root = './Retinex_result/'
path_save = root + 'Denoise_enI_train_DnCNN/'
os.makedirs(path_save, exist_ok=True)
path_save_eval = root + 'Denoise_enI_eval_DnCNN/'
os.makedirs(path_save_eval, exist_ok=True)
Decom_net = DecomNet_RTV(in_ch=1, k1=10)
Decom_net.cuda()
print('====>> load Decom_Net\n')
pre_Decom_checkpoint = './checkpoint/decomnet_V_train_new/checkpoint_99_epoch.pkl'
checkpoint = torch.load(pre_Decom_checkpoint)
Decom_net.load_state_dict(checkpoint['model_state_dict'])
Enhance_I = EnhanceNet_I(in_ch=2, out_ch=1).cuda()
pre_EnhanceI_checkpoint = './checkpoint/Enhancenet_I_train_FD/checkpoint_99_epoch.pkl'
checkpoint = torch.load(pre_EnhanceI_checkpoint)
Enhance_I.load_state_dict(checkpoint['I_model_state_dict'])
Enhance_R = EnhanceNet_I(in_ch=2, out_ch=1).cuda()
pre_EnhanceR_checkpoint = './checkpoint/Enhancenet_VR_train_new/checkpoint_99_epoch.pkl'
checkpoint = torch.load(pre_EnhanceR_checkpoint)
Enhance_R.load_state_dict(checkpoint['R_model_state_dict'])
lr = 1e-5
n_channels = 3
noise_level_img = 15 # set AWGN noise level for noisy image
noise_level_model = noise_level_img # set noise level for model
Denoise_enI = net(in_nc=n_channels + 1, out_nc=n_channels, nc=[64, 128, 256, 512], nb=4, act_mode='R',
downsample_mode="strideconv", upsample_mode="convtranspose").cuda()
Denoise_enI.load_state_dict(torch.load('/home/www/PycharmProjects/myRetinex_new/checkpoint/Denoise_enI_train_new/'
'drunet_color.pth'), strict=True)
Denoise_enI_op = torch.optim.Adam(Denoise_enI.parameters(), lr=lr)
scheduler_enI = torch.optim.lr_scheduler.StepLR(Denoise_enI_op, step_size=100, gamma=0.1) # 设置学习率下降策略
l1 = nn.L1Loss()
mse = nn.MSELoss()
ssim_loss = pytorch_ssim.SSIM()
start_epoch = 0
MAX_epoch = 100
eval_every_epoch = 10
checkpoint_interval = 10
log_interval = 20
train_phase = 'Denoising_enI'
start_time = time.time()
checkpoint_dir = './checkpoint/Denoise_enI_train_DnCNN/'
if not os.path.isdir(checkpoint_dir):
os.makedirs(checkpoint_dir)
ckpt = False
if ckpt:
path_checkpoint = './checkpoint/Denoise_enI_train_DnCNN/checkpoint_9_epoch.pkl'
checkpoint = torch.load(path_checkpoint)
Denoise_enI.load_state_dict(checkpoint['enI_model_state_dict'])
Denoise_enI_op.load_state_dict(checkpoint['enI_optimizer_state_dict'])
# scheduler_H.load_state_dict(checkpoint['H_lr_schedule'])
start_epoch = checkpoint['epoch']
# scheduler_H.last_epoch = start_epoch
for epoch in range(start_epoch, MAX_epoch):
lc_time = time.asctime(time.localtime(time.time()))
Denoise_enI.train()
for i, batch in enumerate(dataloader):
input_low, input_high = Variable(batch[1]), Variable(batch[2], requires_grad=False)
train_low_data = input_low.to(device)
train_high_data = input_high.to(device)
HSV_train_low_data = rgb_to_hsv(train_low_data)
HSV_train_high_data = rgb_to_hsv(train_high_data)
train_low_V = HSV_train_low_data[:, 2].unsqueeze(1)
train_high_V = HSV_train_high_data[:, 2].unsqueeze(1)
low_I_list, low_R_list, alpha, px, py, mu = Decom_net(train_low_V)
high_I_list, high_R_list, alpha, px, py, mu = Decom_net(train_high_V)
low_I, low_R, high_I, high_R = low_I_list[-1], low_R_list[-1], high_I_list[-1], high_R_list[-1]
I_ratio = low_I / (high_I + 0.0001)
enhance_I = Enhance_I(low_I, I_ratio)
enhance_R = Enhance_R(low_R, low_I)
enhance_img = enhance_I * enhance_R
train_low = torch.cat(
[HSV_train_low_data[:, 0].unsqueeze(1), HSV_train_low_data[:, 1].unsqueeze(1), enhance_img], dim=1)
enhance_img = hsv_to_rgb(train_low)
img_L = enhance_img
img_L = torch.cat(
(img_L, torch.FloatTensor([noise_level_model / 255.]).repeat(img_L.shape[0], 1, img_L.shape[2],
img_L.shape[3]).to(device)),
dim=1)
denoise_enI = Denoise_enI(img_L)
Denoise_enI_op.zero_grad()
loss = mse(denoise_enI, train_high_data.detach()) + 1 - ssim_loss(denoise_enI, train_high_data)
loss.backward()
Denoise_enI_op.step()
# for b in range(batch[1].shape[0]):
# name = batch[0][b]
# save_enhance_images(os.path.join(path_save, '%s_%d_%d' % (name, i + 1, epoch + 1)), denoise_enI)
if (i + 1) % log_interval == 0:
print("%s %s Epoch: [%2d] [%4d/%4d] time: %4.4f, loss_I: %.6f" \
% (train_phase, lc_time, epoch + 1, i + 1, numBatch, time.time() - start_time,
loss.data.cpu().numpy()))
scheduler_enI.step() # 更新学习率
if (epoch + 1) % checkpoint_interval == 0:
checkpoint = {"epoch": epoch,
"enI_model_state_dict": Denoise_enI.state_dict(),
"enI_optimizer_state_dict": Denoise_enI_op.state_dict()
# 'H_lr_schedule': scheduler_H.state_dict(),
}
path_checkpoint = checkpoint_dir + "/checkpoint_{}_epoch.pkl".format(epoch)
torch.save(checkpoint, path_checkpoint)
if (epoch + 1) % eval_every_epoch == 0:
print("[*] Evaluating for phase %s / epoch %d..." % (train_phase, epoch + 1))
eval_data = glob.glob(eval_floder[0] + '*.png')
num = len(eval_data)
PSNR = []
SSIM = []
with torch.no_grad():
for path_mat in eval_data:
name = os.path.basename(path_mat)[:-4].split('\\')[-1]
print('validating image:', name)
eval_low = np.asarray(Image.open(eval_floder[0] + name + '.png'))
eval_high = np.asarray(Image.open(eval_floder[1] + name + '.png'))
input_eval_low = Tensor(eval_low).to(device)
input_eval_high = Tensor(eval_high).to(device)
input_eval_low = rgb_to_hsv(input_eval_low)
input_eval_high = rgb_to_hsv(input_eval_high)
eval_low_V = input_eval_low[:, 2].unsqueeze(1)
eval_high_V = input_eval_high[:, 2].unsqueeze(1)
low_I_list, low_R_list, alpha, px, py, mu = Decom_net(eval_low_V)
high_I_list, high_R_list, alpha, px, py, mu = Decom_net(eval_high_V)
# low_I_list, low_R_list, alpha, px, py, mu = Decom_net(input_eval_low)
# high_I_list, high_R_list, alpha, px, py, mu = Decom_net(input_eval_high)
low_I, low_R, high_I, high_R = low_I_list[-1], low_R_list[-1], high_I_list[-1], high_R_list[-1]
I_ratio = low_I / (high_I + 0.0001)
enhance_I = Enhance_I(low_I, I_ratio)
enhance_R = Enhance_R(low_R, low_I)
enhance_img = enhance_I * enhance_R
# input_eval_low[:, 2] = enhance_img
HSV_eval_low = torch.cat(
[input_eval_low[:, 0].unsqueeze(1), input_eval_low[:, 1].unsqueeze(1), enhance_img], dim=1)
enhance_img = hsv_to_rgb(HSV_eval_low)
img_L = enhance_img
img_L = torch.cat(
(img_L,
torch.FloatTensor([noise_level_model / 255.]).repeat(img_L.shape[0], 1, img_L.shape[2],
img_L.shape[3]).to(
device)),
dim=1)
denoise_enI = Denoise_enI(img_L)
enhance_high = denoise_enI[0].cpu().detach().numpy()
enhance_high = np.transpose(enhance_high, (1, 2, 0)) * 255.0
enhance_high = np.clip(enhance_high, 0, 255.0)
psnr_enhance = psnr(enhance_high, eval_high)
ssim_enhance = ssim(enhance_high, eval_high)
print("eval image: %s, PSNR = %5.4f, SSIM = %5.4f" % (name, psnr_enhance, ssim_enhance))
PSNR.append(psnr_enhance)
SSIM.append(ssim_enhance)
save_enhance_images(os.path.join(path_save_eval, '%s_%d_%d' % (name, i + 1, epoch + 1)),
denoise_enI)
avg_PSNR = np.mean(np.asarray(PSNR))
avg_SSIM = np.mean(np.asarray(SSIM))
print("avg_PSNR = %5.4f, avg_SSIM = %5.4f" % (avg_PSNR, avg_SSIM))
## compute PSNR and SSIM
print("[*] Finish training for phase %s." % train_phase)
if __name__ == '__main__':
train_folder = ['/media/www/14F492BBF4929F14/data/LOL/our485/low/', '/media/www/14F492BBF4929F14/data/LOL/our485/high/']
batch_size = 10
train_Data = []
for patch_id in range(batch_size):
rand_mode = np.random.randint(0, 7)
train_data = MyDataset(rand_mode, train_folder)
train_Data.extend(train_data)
print('[*] Number of training data: %d' % len(train_Data))
numBatch = len(train_Data) // int(batch_size)
eval_floder = ['/media/www/14F492BBF4929F14/data/LOL/eval15/low/', '/media/www/14F492BBF4929F14/data/LOL/eval15/high/']
dataloader = DataLoader(dataset=train_Data, batch_size=batch_size, shuffle=True, num_workers=0,
drop_last=True)
train_Denoising_enI(dataloader, eval_floder, numBatch)