-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathtrain_stage3_downstream.py
147 lines (118 loc) · 5.94 KB
/
train_stage3_downstream.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
import torch
import torch.nn.functional as F
from torch.autograd import Variable
import os, argparse
os.environ["CUDA_VISIBLE_DEVICES"] = '0'
from datetime import datetime
from model.model_stage3 import RGBD_sal
from utils_downstream.dataset_rgbd_strategy2 import get_loader
from utils_downstream.utils import adjust_lr, AvgMeter
import torch.nn as nn
from torch.cuda import amp
parser = argparse.ArgumentParser()
parser.add_argument('--epoch', type=int, default=100, help='epoch number')
parser.add_argument('--lr_gen', type=float, default=1e-4, help='learning rate')
parser.add_argument('--batchsize', type=int, default=4, help='training batch size')
parser.add_argument('--trainsize', type=int, default=256, help='training dataset size')
parser.add_argument('--clip', type=float, default=0.5, help='gradient clipping margin')
parser.add_argument('--decay_rate', type=float, default=0.9, help='decay rate of learning rate')
parser.add_argument('--decay_epoch', type=int, default=30, help='every n epochs decay learning rate')
parser.add_argument('-beta1_gen', type=float, default=0.5,help='beta of Adam for generator')
parser.add_argument('--weight_decay', type=float, default=0.001, help='weight_decay')
parser.add_argument('--feat_channel', type=int, default=64, help='reduced channel of saliency feat')
opt = parser.parse_args()
print('Generator Learning Rate: {}'.format(opt.lr_gen))
# build models
generator = RGBD_sal()
generator.cuda()
pretrained_dict = torch.load(os.path.join('./saved_model/pretext_task1.pth'))
model_dict = generator.state_dict()
# print(pretrained_dict.items())
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
model_dict.update(pretrained_dict)
generator.load_state_dict(model_dict)
pretrained_dict = torch.load(os.path.join('./saved_model/pretext_task2.pth'))
model_dict = generator.state_dict()
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
print(pretrained_dict.items())
model_dict.update(pretrained_dict)
generator.load_state_dict(model_dict)
generator_params = generator.parameters()
generator_optimizer = torch.optim.Adam(generator_params, opt.lr_gen)
## load data
image_root = ''
depth_root = ''
gt_root = ''
train_loader = get_loader(image_root, gt_root,depth_root, batchsize=opt.batchsize, trainsize=opt.trainsize)
total_step = len(train_loader)
## define loss
CE = torch.nn.BCELoss()
mse_loss = torch.nn.MSELoss(size_average=True, reduce=True)
# size_rates = [0.75,1,1.25] # multi-scale training
size_rates = [1] # multi-scale training
criterion = nn.BCEWithLogitsLoss().cuda()
criterion_mae = nn.L1Loss().cuda()
criterion_mse = nn.MSELoss().cuda()
use_fp16 = True
scaler = amp.GradScaler(enabled=use_fp16)
def structure_loss(pred, mask):
weit = 1+5*torch.abs(F.avg_pool2d(mask, kernel_size=31, stride=1, padding=15)-mask)
wbce = F.binary_cross_entropy_with_logits(pred, mask, reduction='none')
wbce = (weit*wbce).sum(dim=(2,3))/weit.sum(dim=(2,3))
pred = torch.sigmoid(pred)
inter = ((pred * mask) * weit).sum(dim=(2, 3))
union = ((pred + mask) * weit).sum(dim=(2, 3))
wiou = 1-(inter+1)/(union-inter+1)
return (wbce+wiou).mean()
for epoch in range(1, opt.epoch+1):
generator.train()
loss_record = AvgMeter()
print('Generator Learning Rate: {}'.format(generator_optimizer.param_groups[0]['lr']))
for i, pack in enumerate(train_loader, start=1):
for rate in size_rates:
generator_optimizer.zero_grad()
images, gts, depths = pack
images = Variable(images)
gts = Variable(gts)
depths = Variable(depths)
images = images.cuda()
gts = gts.cuda()
depths = depths.cuda()
# multi-scale training samples
trainsize = int(round(opt.trainsize * rate / 32) * 32)
if rate != 1:
images = F.upsample(images, size=(trainsize, trainsize), mode='bilinear',
align_corners=True)
gts = F.upsample(gts, size=(trainsize, trainsize), mode='bilinear', align_corners=True)
# contours = F.upsample(contours, size=(trainsize, trainsize), mode='bilinear', align_corners=True)
depths = F.upsample(depths, size=(trainsize, trainsize), mode='bilinear', align_corners=True)
b, c, h, w = gts.size()
target_1 = F.upsample(gts, size=h // 2, mode='nearest')
target_2 = F.upsample(gts, size=h // 4, mode='nearest')
target_3 = F.upsample(gts, size=h // 8, mode='nearest')
target_4 = F.upsample(gts, size=h // 16, mode='nearest')
with amp.autocast(enabled=use_fp16):
depth_3 = torch.cat((depths, depths, depths), 1)
sideout5, sideout4, sideout3, sideout2, output1 = generator.forward(images, depth_3) # hed
loss1 = structure_loss(sideout5, target_4)
loss2 = structure_loss(sideout4, target_3)
loss3 = structure_loss(sideout3, target_2)
loss4 = structure_loss(sideout2, target_1)
loss5 = structure_loss(output1, gts)
loss = loss1 + loss2 + loss3 + loss4 + loss5
generator_optimizer.zero_grad()
scaler.scale(loss).backward()
scaler.step(generator_optimizer)
scaler.update()
if rate == 1:
loss_record.update(loss.data, opt.batchsize)
if i % 10 == 0 or i == total_step:
print('{} Epoch [{:03d}/{:03d}], Step [{:04d}/{:04d}], gen Loss: {:.4f}'.
format(datetime.now(), epoch, opt.epoch, i, total_step, loss_record.show()))
# print(anneal_reg)
adjust_lr(generator_optimizer, opt.lr_gen, epoch, opt.decay_rate, opt.decay_epoch)
save_path = './saved_model/SSLSOD_v2'
if not os.path.exists(save_path):
os.makedirs(save_path)
if epoch % opt.epoch == 0:
torch.save(generator.state_dict(), save_path + 'Model' + '_%d' % epoch + '_gen.pth')