-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathtrain.py
184 lines (157 loc) · 6.92 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
"""
description:
version:
Author: zwy
Date: 2023-05-05 15:41:37
LastEditors: zwy
LastEditTime: 2023-05-05 16:25:24
"""
import os
import torch
import torch.nn as nn
import torch.nn.functional as F
import numpy as np
import logging
from datetime import datetime
from tqdm import tqdm
from tensorboardX import SummaryWriter
from torchvision.utils import make_grid
from config import opt
from net import Net
from dataset import get_loader, test_dataset
from loss import MulScaleBoundLoss
from utils import adjust_lr
writer = SummaryWriter(os.path.join(opt.save_path, 'summary/finally'), flush_secs=30)
# logging config
if not os.path.exists(opt.save_path):
os.makedirs(opt.save_path)
logging.basicConfig(filename=os.path.join(opt.save_path, "log", datetime.now().strftime("%Y-%m-%d-%H:%M:%S") + ".log"),
format="[%(asctime)s-%(filename)s-%(levelname)s:%(message)s]",
level=logging.INFO, filemode="a", datefmt="%Y-%m-%d %I:%M:%S %p")
os.environ["CUDA_VISIBLE_DEVICES"] = opt.gpu_id
logging.info("use gpu: {}".format(opt.gpu_id))
# build the model
logging.info("\n<=============build model=============>")
model = Net()
if opt.load is not None:
model.load_state_dict(torch.load(opt.load))
print('load model from ', opt.load)
model.cuda()
params = model.parameters()
optimizer = torch.optim.Adam(params, opt.lr)
logging.info(model)
# load dataset
logging.info("\n<=============load dataset=============>")
image_root = os.path.join(opt.train_root, "RGB/")
ti_root = os.path.join(opt.train_root, "T/")
gt_root = os.path.join(opt.train_root, "GT/")
val_image_root = os.path.join(opt.val_root, "RGB/")
val_ti_root = os.path.join(opt.val_root, "T/")
val_gt_root = os.path.join(opt.val_root, "GT/")
train_loader = get_loader(image_root, gt_root, ti_root, batchsize=opt.batch_size, trainsize=opt.train_size)
test_loader = test_dataset(val_image_root, val_gt_root, val_ti_root, opt.train_size)
total_step = len(train_loader)
logging.info(
"\nepoch:{};lr:{};batch_size:{};train_size:{};clip:{};decay_rate:{};load:{};save_path:{};decay_epoch:{}".format(
opt.epoch, opt.lr, opt.batch_size, opt.train_size, opt.clip, opt.decay_rate, opt.load, opt.save_path,
opt.decay_epoch))
step = 0
best_mae = 1
best_epoch = 0
last_loss = 0
# train function
def train(train_loader, model, optimizer, epoch, save_path):
global step, last_loss
msb_loss = MulScaleBoundLoss()
model.train()
"""
x5_decoder -> torch.Size([1, 256, 40, 30])
x4_decoder -> torch.Size([1, 256, 80, 60])
x3_decoder -> torch.Size([1, 128, 160, 120])
x2_decoder -> torch.Size([1, 64, 320, 240])
out -> torch.Size([1, 9, 640, 480])
"""
loss_all = 0
epoch_step = 0
last_loss = loss_all
try:
for i, (images, gts, tis) in enumerate(train_loader, start=1):
optimizer.zero_grad()
rgb = images.cuda()
ir = tis.cuda()
gt = gts.cuda()
# print("gt shape:", gt.shape)
out = model(rgb, ir)
loss, bound_i, pre_bound_i = msb_loss(gt, out)
loss.backward()
optimizer.step()
step = step + 1
epoch_step = epoch_step + 1
loss_all = loss.item() + loss_all
if i % 10 == 0 or i == total_step or i == 1:
print("{} Epoch [{:03d}/{:03d}], Step [{:04d}/{:04d}], Loss: {:.4f}, loss_all : {:.4f}".
format(datetime.now().strftime("%Y-%m-%d-%H:%M:%S"), epoch, opt.epoch, i, total_step,
loss.item(), loss_all))
logging.info("{} Epoch [{:03d}/{:03d}], Step [{:04d}/{:04d}], Loss: {:.4f}, loss_all : {:.4f}".
format(datetime.now().strftime("%Y-%m-%d-%H:%M:%S"), epoch, opt.epoch, i, total_step, loss.item(), loss_all))
writer.add_scalar('Loss', loss, global_step=step)
grid_image = make_grid(gts[0].clone().cpu().data, 1, normalize=True)
writer.add_image('train/Ground_truth', grid_image, step)
grid_image = make_grid(bound_i.clone().cpu().data, 1, normalize=True)
writer.add_image('train/bound', grid_image, step)
res = out[0][0].clone()
res = res.sigmoid().data.cpu().numpy().squeeze()
res = (res - res.min()) / (res.max() - res.min() + 1e-8)
writer.add_image('OUT/out', torch.tensor(res), step, dataformats='HW')
res = pre_bound_i.clone()
res = res.sigmoid().data.cpu().numpy().squeeze()
res = (res - res.min()) / (res.max() - res.min() + 1e-8)
writer.add_image('OUT/bound', torch.tensor(res), step, dataformats='HW')
loss_all /= epoch_step
writer.add_scalar('Loss-epoch', loss_all, global_step=epoch)
torch.save(model.state_dict(), os.path.join(save_path, "models/finally", 'MSEDNET_Last.pth'))
except KeyboardInterrupt:
print('Keyboard Interrupt: save model and exit.')
if not os.path.exists(save_path):
os.makedirs(save_path)
torch.save(model.state_dict(), os.path.join(save_path, "models/finally", 'MSEDNET_Inter.pth'))
print('save checkpoints successfully!')
raise
# test function
def test(test_loader, model, epoch, save_path):
global best_mae, best_epoch
model.eval()
with torch.no_grad():
mae_sum = 0
for i in tqdm(range(test_loader.size)):
image, gt, ti, name = test_loader.load_data()
gt = gt.cuda()
image = image.cuda()
ti = ti.cuda()
res = model(image, ti)
res = torch.sigmoid(res)
res = (res - res.min()) / (res.max() - res.min() + 1e-8)
mae_train = torch.sum(torch.abs(res - gt)) * 1.0 / (torch.numel(gt))
# print(mae_train)
mae_sum = mae_train.item() + mae_sum
# print(test_loader.size)
mae = mae_sum / test_loader.size
# print(test_loader.size)
writer.add_scalar('MAE', torch.as_tensor(mae), global_step=epoch)
print('Epoch: {} MAE: {} #### bestMAE: {} bestEpoch: {}'.format(epoch, mae, best_mae, best_epoch))
if epoch == 1:
best_mae = mae
else:
if mae < best_mae:
best_mae = mae
best_epoch = epoch
torch.save(model.state_dict(), os.path.join(save_path, "models/finally", 'MSEDNET_Best.pth'))
print('best epoch:{}'.format(epoch))
logging.info('#TEST#:Epoch:{} MAE:{} bestEpoch:{} bestMAE:{}'.format(epoch, mae, best_epoch, best_mae))
if __name__ == '__main__':
print("Start train...")
for epoch in range(1, opt.epoch + 1):
cur_lr = adjust_lr(optimizer, opt.lr, epoch, opt.decay_rate, opt.decay_epoch)
writer.add_scalar('learning_rate', cur_lr, global_step=epoch)
train(train_loader, model, optimizer, epoch, opt.save_path)
test(test_loader, model, epoch, opt.save_path)