-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
399 lines (325 loc) · 14.7 KB
/
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
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
394
395
396
397
398
399
import argparse
import os
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader
from networks import define_G, define_D, GANLoss, print_network, get_conv
from dataset import PanColorDataset, PanSRDataset
import torch.backends.cudnn as cudnn
from util import save_figure, visualize_tensor, avg_metric
import numpy as np
import matplotlib.pyplot as plt
from metrics import sCC
from metrics import ERGAS as ergas
from metrics import sam2 as sam
import torch.multiprocessing
torch.multiprocessing.set_sharing_strategy('file_system')
# Training settings
parser = argparse.ArgumentParser(description='PanColorGAN-PyTorch-implementation')
#parser.add_argument('--datasetSource', required=True, help='path of dataset')
#parser.add_argument('--datasetTarget', required=False, help='path of dataset')
parser.add_argument('--dataPath', help='path of data')
parser.add_argument('--dataset', type=str, default='pleiades')
parser.add_argument('--savePath', required=True, help='path of save')
parser.add_argument('--batchSize', type=int, default=16,
help='training batch size')
parser.add_argument('--nEpochs', type=int, default=200,
help='number of epochs to train for')
parser.add_argument('--testEveryNEpochs', type=int, default=2,
help='Test every n epochs')
parser.add_argument('--model', type=str, default='PanColorGAN')
parser.add_argument("--useRD", action='store_true')
parser.add_argument('--input_nc', type=int, default=5,
help='input image channels')
parser.add_argument('--output_nc', type=int, default=4,
help='output image channels')
parser.add_argument('--ngf', type=int, default=64,
help='generator filters in first conv layer')
parser.add_argument('--ndf', type=int, default=64,
help='discriminator filters in first conv layer')
parser.add_argument('--nlayers', type=int, default=5)
parser.add_argument('--nblocks', type=int, default=6)
parser.add_argument('--ndowns', type=int, default=2)
parser.add_argument('--gtype', type=str, default='fusenet')
parser.add_argument('--lr', type=float, default=0.0002,
help='Learning Rate. Default=0.0002')
parser.add_argument('--adjustLR', action='store_true',
help='decrease LR by %1 after each epoch 100th')
parser.add_argument('--beta1', type=float, default=0.5,
help='beta1 for adam. default=0.5')
parser.add_argument('--cuda', action='store_true', help='use cuda?')
parser.add_argument('--threads', type=int, default=4,
help='number of threads for data loader to use')
parser.add_argument('--seed', type=int, default=123,
help='random seed to use. Default=123')
parser.add_argument('--lamb', type=float, default=1,
help='weight on L1 term in objective')
parser.add_argument('--weightG', type=float, default=1)
parser.add_argument("--useL2", action='store_true')
parser.add_argument('--useDropout', action='store_true')
parser.add_argument("--lsgan", action='store_true', default=False,
help='use lsgan loss in D')
parser.add_argument('--upConvType', type=str, default='ConvT',
help='type of upsampling conv, default is ConvTranspose2d')
parser.add_argument('--blockType', type=str, default='SE',
help='Type of Generator block (Resnet, RRDB)')
parser.add_argument('--lossType', type=str, default='ragan',
help='Type of GAN Loss (Normal, ragan(relativistic average gan)')
parser.add_argument('--cont', action='store_true', help='continue from')
parser.add_argument('--checkpointPath')
parser.add_argument('--contEpoch', type=int,
help='contiune from where we left', default=0)
parser.add_argument("--hddPath", type=str, default='./')
parser.add_argument("--regTerm", type=float, default=0.0)
parser.add_argument('--gpuSet', type=int, default=1)
opt = parser.parse_args()
print(opt)
if not os.path.exists("results-{}".format(opt.savePath)):
os.mkdir("results-{}".format(opt.savePath))
f = open('results-{}/metrics.txt'.format(opt.savePath), 'w+')
F = open("results-{}/params.txt".format(opt.savePath), 'w')
# F.write(str(opt).split())
for i in str(opt).split():
F.write(i + '\n')
F.close()
f.close()
## GPU Initialization
if opt.cuda and not torch.cuda.is_available():
raise Exception("No GPU found, please run without --cuda")
cudnn.benchmark = True
gpus = [gpu for gpu in range(opt.gpuSet)]
torch.cuda.set_device(gpus[0])
## Data Initialization
if opt.model == 'PanColorGAN':
train_set = PanColorDataset(mode='train', dataset=opt.dataset, random_downsampling=opt.useRD)
test_set = PanColorDataset(mode='test', dataset=opt.dataset)
elif opt.model == 'PanSRGAN':
train_set = PanSRDataset(mode='train', dataset=opt.dataset)
test_set = PanSRDataset(mode='test', dataset=opt.dataset)
training_data_loader = DataLoader(
dataset=train_set, num_workers=opt.threads, batch_size=opt.batchSize, shuffle=True)
testing_data_loader = DataLoader(
dataset=test_set,num_workers=opt.threads, batch_size=opt.batchSize, shuffle=False)
## Define Network
netG = define_G(opt.input_nc, opt.output_nc,
opt.ngf, 'batch','leakyrelu', opt.useDropout, opt.upConvType, opt.gtype, opt.blockType, opt.nblocks, gpus, n_downsampling=opt.ndowns)
netD = define_D(opt.input_nc + opt.output_nc,
opt.ndf, 'batch', not opt.lsgan, opt.nlayers, gpus)
## Define Losses
criterionGAN = GANLoss(use_lsgan=opt.lsgan)
criterionL1 = nn.L1Loss()
criterionMSE = nn.MSELoss()
## Define Optimizers
optimizerG = optim.Adam(netG.parameters(), lr=opt.lr, betas=(
opt.beta1, 0.999), weight_decay=opt.regTerm)
optimizerD = optim.Adam(netD.parameters(), lr=opt.lr, betas=(
opt.beta1, 0.999), weight_decay=opt.regTerm)
## Continue from a checkpoint
if opt.cont:
## Load Networks
netG.load_state_dict(torch.load(
'checkpoint/{}/netG_model_epoch_{}.pth'.format(opt.savePath, opt.contEpoch),
map_location=lambda storage, loc: storage))
netD.load_state_dict(torch.load(
'checkpoint/{}/netD_model_epoch_{}.pth'.format(opt.savePath, opt.contEpoch),
map_location=lambda storage, loc: storage))
## Load Optimizers
optimizerG.load_state_dict(
torch.load('{}checkpoint/{}/optimG_model_epoch_{}.pth'.format(opt.hddPath, opt.savePath, opt.contEpoch), map_location=lambda storage, loc: storage))
for state in optimizerG.state.values():
for k, v in state.items():
if isinstance(v, torch.Tensor):
state[k] = v.cuda()
optimizerD.load_state_dict(
torch.load('{}checkpoint/{}/optimD_model_epoch_{}.pth'.format(opt.hddPath, opt.savePath, opt.contEpoch), map_location=lambda storage, loc: storage))
for state in optimizerD.state.values():
for k, v in state.items():
if isinstance(v, torch.Tensor):
state[k] = v.cuda()
opt.nEpochs += opt.contEpoch
## Loss Lists Initialization
losses_dict = {"batchId": {}, "d": {}, "g": {},
"grad": {}, "recon": {}, "adv": {}}
losses_g = []
losses_d = []
losses_recon = []
losses_adv = []
losses_recon_epoch = []
losses_adv_epoch = []
losses_g_epoch = []
losses_d_epoch = []
## Tensors Initialized
real_a = torch.FloatTensor(opt.batchSize, opt.input_nc, 256, 256)
real_b = torch.FloatTensor(opt.batchSize, opt.output_nc, 256, 256)
## Switch to CUDA tensors
if opt.cuda:
netD = netD.cuda()
netG = netG.cuda()
netD.train()
netG.train()
criterionGAN = criterionGAN.cuda()
criterionL1 = criterionL1.cuda()
criterionMSE = criterionMSE.cuda()
real_a = real_a.cuda()
real_b = real_b.cuda()
def train(epoch):
netG.train()
netD.train()
for iteration, batch in enumerate(training_data_loader, 1):
real_a_cpu, real_b_cpu = batch[0].view(
-1, opt.input_nc, 256, 256), batch[1].view(-1, opt.output_nc, 256, 256)
real_a.data.resize_(real_a_cpu.size()).copy_(real_a_cpu)
real_b.data.resize_(real_b_cpu.size()).copy_(real_b_cpu)
losses_dict["batchId"]["{}-{}".format(epoch, iteration)] = batch[2]
fake_b = netG(real_a)
############################
# (1) Update D network: maximize log(D(x,y)) + log(1 - D(x,G(x)))
###########################
optimizerD.zero_grad()
# train with fake
fake_ab = torch.cat((real_a, fake_b), 1)
pred_fake= netD.forward(fake_ab.detach())
# train with real
real_ab = torch.cat((real_a, real_b), 1)
pred_real = netD.forward(real_ab)
# Relativistic Average GAN Loss or Normal GAN Loss
if opt.lossType == 'ragan':
loss_d_fake = criterionGAN(
pred_fake - torch.mean(pred_real), False)
loss_d_real = criterionGAN(pred_real - torch.mean(pred_fake), True)
elif opt.lossType == 'gan':
loss_d_fake = criterionGAN(pred_fake, False)
loss_d_real = criterionGAN(pred_real, True)
del fake_ab
del pred_fake
del real_ab
del pred_real
# Combined loss
loss_d = (loss_d_fake + loss_d_real) * 0.5
loss_d.backward()
losses_d.append(loss_d.data.cpu().item())
losses_dict["d"]["{}-{}".format(epoch, iteration)
] = loss_d.data.cpu().item()
del loss_d_fake
del loss_d_real
del loss_d
# Parameter optimizing
nn.utils.clip_grad_norm_(netD.parameters(), 0.3)
nn.utils.clip_grad_norm_(netG.parameters(), 0.3)
optimizerD.step()
############################
# (2) Update G network: maximize log(D(x,G(x))) + L1(y,G(x))
##########################
optimizerG.zero_grad()
# First, G(A) should fake the discriminator
if opt.lossType == 'ragan':
# train with fake
fake_ab = torch.cat((real_a, fake_b), 1)
pred_fake = netD.forward(fake_ab)
# train with real
real_ab = torch.cat((real_a, real_b), 1)
pred_real = netD.forward(real_ab)
loss_g_gan = (criterionGAN(pred_real - torch.mean(pred_fake), False) +
criterionGAN(pred_fake - torch.mean(pred_real), True)) / 2
del fake_ab
del pred_fake
del real_ab
del pred_real
elif opt.lossType == 'gan':
fake_ab = torch.cat((real_a, fake_b), 1)
pred_fake = netD.forward(fake_ab)
loss_g_gan = criterionGAN(pred_fake, True)
del fake_ab
del pred_fake
# Second, G(A) = B
loss_g_l1 = (criterionL1(fake_b, real_b) ) * opt.lamb
loss_g = (loss_g_gan * opt.weightG) + loss_g_l1
loss_g.backward()
losses_g.append(loss_g.data.cpu().item())
losses_adv.append(loss_g_gan.data.cpu().item())
losses_recon.append(loss_g_l1.data.cpu().item() * 1 / opt.lamb)
losses_dict["g"]["{}-{}".format(epoch, iteration)
] = loss_g.data.cpu().item()
losses_dict["adv"]["{}-{}".format(epoch,
iteration)] = loss_g_gan.data.cpu().item()
losses_dict["recon"]["{}-{}".format(epoch, iteration)] = loss_g_l1.data.cpu().item() * 1 / opt.lamb
del loss_g
del loss_g_l1
del loss_g_gan
# Parameter optimizing
# nn.utils.clip_grad_norm(netD.parameters(), 0.3)
# nn.utils.clip_grad_norm(netG.parameters(), 0.3)
optimizerG.step()
#torch.cuda.synchronize()
print("===> Epoch[{}]({}/{}): Loss_D: {:.4f} Loss_G: {:.4f}, Loss_Recon: {:.7f} ".format(
epoch, iteration, len(training_data_loader), losses_d[-1], losses_g[-1], losses_recon[-1]))
del fake_b
len_loader = len(training_data_loader)
losses_d_epoch.append(
np.mean(losses_d[-len_loader:]))
losses_g_epoch.append(
np.mean(losses_g[-len_loader:]))
losses_adv_epoch.append(
np.mean(losses_adv[-len_loader:]))
losses_recon_epoch.append(
np.mean(losses_recon[-len_loader:]))
def test(epoch):
netG.eval()
if not os.path.isdir("results-{}/test".format(opt.savePath)):
os.mkdir("results-{}/test".format(opt.savePath))
avg_sCC = []
avg_sam = []
avg_ergas = []
for it, batch in enumerate(testing_data_loader):
input,target= batch[0],batch[1]
input = input.view(-1, opt.input_nc, 256, 256)
target = target.cuda().view(-1, opt.output_nc, 256, 256)
prediction = netG(input.view(-1, opt.input_nc, 256, 256).cuda())
sam_val = avg_metric(target, prediction, sam)
avg_sam.append(sam_val)
sCC_val = avg_metric(target, prediction, sCC)
avg_sCC.append(sCC_val)
ergas_val = avg_metric(target, prediction, ergas)
avg_ergas.append(ergas_val)
del input
del target
del prediction
del sCC_val
del sam_val
del ergas_val
res = "===> Avg. SAM: {:.4f} , Avg sCC: {:.4f}, Avg ERGAS: {:.4f} epoch: {} \n".format(
np.mean(avg_psnr), np.std(avg_psnr), np.mean(avg_sam), np.mean(avg_sCC), np.mean(avg_ergas), epoch)
f = open('results-{}/metrics.txt'.format(opt.savePath), 'a')
f.write(res)
print(res)
def checkpoint(epoch):
if opt.hddPath is not None:
path = os.path.join(opt.hddPath, "checkpoint")
else:
path = "checkpoint"
if not os.path.exists(path):
os.mkdir(path)
if not os.path.exists(os.path.join(path, opt.savePath)):
os.mkdir(os.path.join(path, opt.savePath))
net_g_model_out_path = "{}/{}/netG_model_epoch_{}.pth".format(path,
opt.savePath, epoch)
net_d_model_out_path = "{}/{}/netD_model_epoch_{}.pth".format(path,
opt.savePath, epoch)
optim_d_model_out_path = "{}/{}/optimD_model_epoch_{}.pth".format(path,
opt.savePath, epoch)
optim_g_model_out_path = "{}/{}/optimG_model_epoch_{}.pth".format(path,
opt.savePath, epoch)
torch.save(netG.state_dict(), net_g_model_out_path)
torch.save(netD.state_dict(), net_d_model_out_path)
torch.save(optimizerD.state_dict(), optim_d_model_out_path)
torch.save(optimizerG.state_dict(), optim_g_model_out_path)
print("Checkpoint saved to {}".format(path + opt.savePath))
for epoch in range(opt.contEpoch, opt.nEpochs + 1):
train(epoch)
if epoch % opt.testEveryNEpochs == 0:
checkpoint(epoch)
print('===> Testing')
test(epoch)
np.savez("results-{}/losses".format(opt.savePath), np.array(losses_dict))
f.close()