-
Notifications
You must be signed in to change notification settings - Fork 32
/
Copy pathStep3_Pretrain_Unet.py
184 lines (143 loc) · 5.45 KB
/
Step3_Pretrain_Unet.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Fri Feb 19 14:41:32 2021
@author: li0005
"""
import torch
import torch.optim as optim
import util.util as util
import csv
import os
import argparse
from datetime import date
import UNet.unet_load_data as load_dataset
import UNet.UNet as unet
import torch.nn.functional as F
import time
par = argparse.ArgumentParser(description='UNet')
par.add_argument('--epochs',default=51,type=int,help='Total epochs')
par.add_argument('--batch_size',default=12,type=int,help='Batch sizes')
par.add_argument('--gpu',default=0,type=int,help='The GPU ID')
par.add_argument('--pretrained_model_train',default=0,type=int,help='Pretrained model')
par.add_argument('--img_path',type=str,help='Root of the training samples')
current_path = os.getcwd()
args = par.parse_args()
batch = args.batch_size
epoch = args.epochs
GPU_no = args.gpu
ct = args.pretrained_model_train #load trained model
output_name = 'Pretrain_UNet'
device = torch.device("cuda:{}".format(util.device_ids[GPU_no]) if torch.cuda.is_available() else "cpu")
image_path = (args.img_path + '/' ).replace('//','/')
current_path = os.getcwd()
output_path = current_path+'/MoFA_UNet_Save/'+output_name + '/'
if not os.path.exists(output_path):
os.makedirs(output_path)
width = 224
height = 224
'''----------------
#load images
----------------'''
trainset = load_dataset.UNetDataset(device,image_path, True,height,width,1)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch,shuffle=True, num_workers=0)
testset = load_dataset.UNetDataset(device,image_path,False,height,width,1)
testloader = torch.utils.data.DataLoader(testset, batch_size=batch,shuffle=False, num_workers=0)
images_org_test = []
images_mask_test = []
images_rasters_test = []
images_input = []
test_batch_num = 3
for i_test, data_test in enumerate(testloader, 0):
if i_test >= test_batch_num:
break
inputs_test, masks_test, orgs_test, rasters_test= data_test
images_mask_test += [masks_test]
images_input += [inputs_test]
util.write_tiled_image(torch.cat(images_mask_test,dim=0), output_path + 'test_mask_gt.png',10)
'''----------------
Prepare log files
----------------'''
today = date.today()
loss_log_path_train = output_path+today.strftime("%b-%d-%Y")+"loss_train.csv"
loss_log_path_test = output_path+today.strftime("%b-%d-%Y")+"loss_test.csv"
#prepare log file
fid_train = open(loss_log_path_train, 'w')
writer_train = csv.writer(fid_train, lineterminator="\r\n")
fid_test = open(loss_log_path_test, 'w')
writer_test = csv.writer(fid_test, lineterminator="\r\n")
'''----------------
Prepare Network and Optimizer
----------------'''
unet_mask = unet.UNet(in_ch=6, out_ch=1).to(device)
'''----------------
UNet Forward
----------------'''
def proc_unet(input_imgs,mask_gt):
pred_mask = unet_mask(input_imgs)
cross_entropy_loss = F.binary_cross_entropy_with_logits(pred_mask, mask_gt)
return cross_entropy_loss,pred_mask
if ct!=0:
trained_model_path = output_path+'unet_mask_{:06d}.model'.format(ct)
unet_mask = torch.load(trained_model_path, map_location='cuda:{}'.format(util.device_ids[GPU_no]))
print('Loading pre-trained model:'+ output_path + ' unet_mask_{:06d}.model'.format(ct))
optimizer = optim.Adadelta(unet_mask.parameters(), lr=0.03)
scheduler = optim.lr_scheduler.StepLR(optimizer,step_size=5000,gamma=0.99)
start = time.time()
mean_losses = torch.zeros([1])
for ep in range(0,epoch):
for i, data in enumerate(trainloader, 0):
#Evaluation
if ct % 1000 == 0 and ct>0:
test_masks = []
for images_temp, mask_gt_temp in zip(images_input , images_mask_test):
with torch.no_grad():
unet_mask.eval()
loss_, test_mask_temp = proc_unet(images_temp, mask_gt_temp)
test_masks += [test_mask_temp]
util.write_tiled_image(torch.cat(test_masks,dim=0),output_path+'test_image_{}.png'.format(ct),10)
if ct % 5000 ==0:
torch.save(unet_mask, output_path + 'unet_mask_{:06d}.model'.format(ct))
#validating
if ct % 2000 == 0:
print('UNet Training mode:'+output_name)
c_test=0
mean_test_losses = torch.zeros([1])
unet_mask.eval()
for i_test, data_test in enumerate(testloader,0):
image_input_temp, mask_gt_temp, _, _ = data_test
c_test+=1
with torch.no_grad():
loss_, est_mask_temp = proc_unet(image_input_temp, mask_gt_temp)
mean_test_losses += torch.FloatTensor([loss_.item()])
mean_test_losses = mean_test_losses/c_test
str = 'validation loss:{} {:.5f}'.format(ct, mean_test_losses[0])
print(str)
writer_test.writerow(str)
fid_train.close()
fid_train = open(loss_log_path_train , 'a')
writer_train = csv.writer(fid_train, lineterminator="\r\n")
fid_test.close()
fid_test = open(loss_log_path_test, 'a')
writer_test = csv.writer(fid_test, lineterminator="\r\n")
#Training
unet_mask.train()
image_train, mask_gt_train, _,_ = data
if image_train.shape[0]!=batch:
continue
optimizer.zero_grad()
loss, est_mask = proc_unet(image_train, mask_gt_train)
mean_losses += torch.FloatTensor([loss.item()])
loss.backward()
optimizer.step()
if ct % 100 == 0 and ct>0:
end = time.time()
mean_losses = mean_losses/100
str = 'train loss:{} {:.5f} {}'.format(ct, mean_losses[0],end-start)
start = end
print(str)
writer_train.writerow(str)
mean_losses = torch.zeros([1])
scheduler.step()
ct += 1
torch.save(unet_mask, output_path + 'unet_mask_{:06d}.model'.format(ct))