-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathtrain.py
More file actions
205 lines (182 loc) · 9.68 KB
/
train.py
File metadata and controls
205 lines (182 loc) · 9.68 KB
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
import argparse
import time
from torch.autograd import Variable
from torch.utils.data import DataLoader
from net import Net
from dataset import *
import matplotlib.pyplot as plt
from metrics import *
import numpy as np
import os
from torch.backends import cudnn
os.environ['KMP_DUPLICATE_LIB_OK'] = 'TRUE'
parser = argparse.ArgumentParser(description="PyTorch BasicIRSTD train")
parser.add_argument("--model_names", default=['MTU'], type=list,
help="model_name: 'DNANet', 'ISNet', 'UIUNet', 'RDIAN', 'U-Net', 'SCTransNet', 'MTU', 'MSH', 'APT', 'ILNet'")
parser.add_argument("--conv", default='DHiF', type=str, help="convolution types: usual, DHiF, WTC, SDC, PC, FD, Ref")
parser.add_argument("--dataset_names", default=['NUAA-SIRST'], type=list,
help="dataset_name: 'NUAA-SIRST', 'IRSTD-1K', 'SIRST3', 'NUDT-SIRST-Sea'")
parser.add_argument("--img_norm_cfg", default=None, type=dict,
help="specific a img_norm_cfg, default=None (using img_norm_cfg values of each dataset)")
parser.add_argument("--dataset_dir", default='./dataset', type=str, help="train_dataset_dir")
parser.add_argument("--batchSize", type=int, default=16, help="Training batch size")
parser.add_argument("--seed", type=int, default=1337, help="Random seed")
parser.add_argument("--grad-accum-steps", type=int, default=1, help="Gradient accumulation steps")
parser.add_argument("--patchSize", type=int, default=256, help="Training patch size")
parser.add_argument("--interference", default=False, type=bool, help="Dataset Augment")
parser.add_argument("--save", default='./log_new', type=str, help="Save path of checkpoints")
parser.add_argument("--resume", default=None, type=list, help="Resume from exisiting checkpoints (default: None)")
parser.add_argument("--nEpochs", type=int, default=500, help="Number of epochs")
parser.add_argument("--optimizer_name", default='Adam', type=str, help="optimizer name: Adam, Adagrad, SGD")
parser.add_argument("--optimizer_settings", default={'lr': 5e-4}, type=dict, help="optimizer settings")
parser.add_argument("--scheduler_name", default='MultiStepLR', type=str, help="scheduler name: MultiStepLR")
parser.add_argument("--scheduler_settings", default={'step': [100, 150, 200, 250, 300, 350, 400, 450, 500], 'gamma': 0.5}, type=dict, help="scheduler settings")
parser.add_argument("--threads", type=int, default=0, help="Number of threads for data loader to use")
parser.add_argument("--threshold", type=float, default=0.5, help="Threshold for test")
global opt
opt = parser.parse_args()
cuda = '0'
os.environ["CUDA_VISIBLE_DEVICES"] = cuda
mark = '_500epoch_' + opt.conv
print('cuda:', cuda)
def init_seeds(seed=0, cuda_deterministic=True):
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
os.environ['PYTHONHASHSEED'] = str(seed)
# Speed-reproducibility tradeoff https://pytorch.org/docs/stable/notes/randomness.html
if cuda_deterministic: # slower, more reproducible
cudnn.deterministic = True
cudnn.benchmark = False
# torch.use_deterministic_algorithms(True)
else: # faster, less reproducible
cudnn.deterministic = False
cudnn.benchmark = True
# torch.use_deterministic_algorithms(False)
def train():
interference = opt.interference
if not interference:
train_set = TrainSetLoader(dataset_dir=opt.dataset_dir, dataset_name=opt.dataset_name, patch_size=opt.patchSize, img_norm_cfg=opt.img_norm_cfg)
else:
train_set = TrainSetLoader_aug(dataset_dir=opt.dataset_dir, dataset_name=opt.dataset_name, patch_size=opt.patchSize, img_norm_cfg=opt.img_norm_cfg)
total_loss_pred_epoch = []
train_loader = DataLoader(dataset=train_set, num_workers=opt.threads, batch_size=opt.batchSize, shuffle=True, pin_memory=False, persistent_workers=False)
save_path = opt.save + '/' + opt.dataset_name + '/' + opt.model_name + mark
os.makedirs(save_path, exist_ok=True)
net = Net(model_name=opt.model_name, mode='train', conv=opt.conv, save_path=save_path).cuda()
net.train()
epoch_state = 0
total_loss_list = []
total_loss_epoch = []
if opt.resume:
for resume_pth in opt.resume:
if opt.dataset_name in resume_pth and opt.model_name in resume_pth:
ckpt = torch.load(resume_pth)
net.load_state_dict(ckpt['state_dict'])
epoch_state = ckpt['epoch']
total_loss_list = ckpt['total_loss']
# for i in range(len(opt.step)):
# opt.step[i] = opt.step[i] - ckpt['epoch']
### Default settings of DNANet
if opt.optimizer_name == 'Adagrad':
opt.optimizer_settings['lr'] = 0.05
opt.scheduler_name = 'CosineAnnealingLR'
opt.scheduler_settings['epochs'] = 1500
opt.scheduler_settings['min_lr'] = 1e-3
opt.nEpochs = opt.scheduler_settings['epochs']
optimizer, scheduler = get_optimizer(net, opt.optimizer_name, opt.scheduler_name, opt.optimizer_settings, opt.scheduler_settings)
for idx_epoch in range(epoch_state, opt.nEpochs):
for idx_iter, (img, gt_mask) in enumerate(train_loader):
img, gt_mask = Variable(img).cuda(), Variable(gt_mask).cuda()
if img.shape[0] == 1:
continue
if not interference:
pred = net.forward(img, idx_epoch)
loss = net.loss(pred, gt_mask)
else:
pred = net.forward(img[:,0,:,:].unsqueeze(1), idx_epoch)
pred_interf = net.forward(img[:,1,:,:].unsqueeze(1), idx_epoch)
loss, loss_pred = net.loss_interf(pred, pred_interf, gt_mask)
total_loss_pred_epoch.append(loss_pred.detach().cpu())
total_loss_epoch.append(loss.detach().cpu())
loss = loss / opt.grad_accum_steps
loss.backward()
if (idx_iter + 1) % opt.grad_accum_steps == 0 or (idx_iter + 1) == len(train_loader):
optimizer.step()
optimizer.zero_grad()
scheduler.step()
if (idx_epoch + 1) % 10 == 0:
total_loss_list.append(float(np.array(total_loss_epoch).mean()))
if not interference:
print(time.ctime()[4:-5] + ' Epoch---%d, total_loss---%f,'
% (idx_epoch + 1, total_loss_list[-1]))
opt.f.write(time.ctime()[4:-5] + ' Epoch---%d, total_loss---%f,\n'
% (idx_epoch + 1, total_loss_list[-1]))
total_loss_epoch = []
else:
print(time.ctime()[4:-5] + ' Epoch---%d, total_loss---%f, total_loss_pred---%f,'
% (idx_epoch + 1, total_loss_list[-1], float(np.array(total_loss_pred_epoch).mean())))
opt.f.write(time.ctime()[4:-5] + ' Epoch---%d, total_loss---%f, total_loss_pred---%f,\n'
% (idx_epoch + 1, total_loss_list[-1], float(np.array(total_loss_pred_epoch).mean())))
total_loss_epoch = []
total_loss_pred_epoch = []
if (idx_epoch + 1) % 50 == 0 and (idx_epoch + 1) != opt.nEpochs:
save_pth = save_path + '/' + str(idx_epoch + 1) + '.pth.tar' #
save_checkpoint({
'epoch': idx_epoch + 1,
'state_dict': net.state_dict(),
'total_loss': total_loss_list,
}, save_pth)
test(save_pth)
if (idx_epoch + 1) == opt.nEpochs:
save_pth = save_path + '/' + str(idx_epoch + 1) + '.pth.tar' #
save_checkpoint({
'epoch': idx_epoch + 1,
'state_dict': net.state_dict(),
'optim_dict': optimizer.state_dict(),
'total_loss': total_loss_list,
}, save_pth)
test(save_pth)
def test(save_pth):
test_set = TestSetLoader(opt.dataset_dir, opt.dataset_name, opt.dataset_name, img_norm_cfg=opt.img_norm_cfg)
test_loader = DataLoader(dataset=test_set, num_workers=1, batch_size=1, shuffle=False)
net = Net(model_name=opt.model_name, mode='test', conv=opt.conv).cuda()
ckpt = torch.load(save_pth)
net.load_state_dict(ckpt['state_dict'])
net.eval()
eval_mIoU = mIoU()
eval_PD_FA = PD_FA()
for idx_iter, (img, gt_mask, size, _) in enumerate(test_loader):
img = Variable(img).cuda()
pred = net.forward(img, 100)
pred = pred[:,:,:size[0],:size[1]]
gt_mask = gt_mask[:,:,:size[0],:size[1]]
eval_mIoU.update((pred>opt.threshold).cpu(), gt_mask)
eval_PD_FA.update((pred[0,0,:,:]>opt.threshold).cpu(), gt_mask[0,0,:,:], size)
results1 = eval_mIoU.get()
results2 = eval_PD_FA.get()
print("IoU, nIoU:\t" + str(results1))
print("PD, FA:\t" + str(results2))
opt.f.write("IoU, nIoU:\t" + str(results1) + '\n')
opt.f.write("PD, FA:\t" + str(results2) + '\n')
def save_checkpoint(state, save_path):
if not os.path.exists(os.path.dirname(save_path)):
os.makedirs(os.path.dirname(save_path))
torch.save(state, save_path)
return save_path
if __name__ == '__main__':
if opt.seed is not None:
init_seeds(opt.seed)
for dataset_name in opt.dataset_names:
opt.dataset_name = dataset_name
for model_name in opt.model_names:
opt.model_name = model_name
if not os.path.exists(opt.save):
os.makedirs(opt.save)
opt.f = open(opt.save + '/' + opt.dataset_name + '_' + opt.model_name + mark + '_' + (time.ctime()).replace(' ', '_').replace(':', '-') + '.txt', 'w')
print(opt.dataset_name + '\t' + opt.model_name + mark)
train()
print('\n')
opt.f.close()