-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrainNet1.py
185 lines (164 loc) · 6.64 KB
/
trainNet1.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
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader
import torchvision
import torchvision.transforms as transforms
import numpy as np
import os
import random
import cv2
import argparse
from PIL import Image
from ResFCN256 import ResFCN256
from load300wlp import Dataset300WLP, ToTensor, ToNormalize
'''''
class LoadTrainData(object):
def __init__(self, train_data_file):
super(LoadTrainData, self).__init__()
self.train_data_file = train_data_file
self.train_data_list = []
self.readTrainData()
self.index = 0
self.num_data = len(self.train_data_list)
def readTrainData(self):
with open(self.train_data_file) as fp:
temp = fp.readlines()
for item in temp:
item = item.strip().split()
self.train_data_list.append(item)
random.shuffle(self.train_data_list)
def getBatch(self, batch_list):
batch = []
imgs = []
labels = []
for item in batch_list:
if len(item)==2:
img_name = item[0]
label_name = item[1]
else:
img_name = item[0] + ' ' + item[1]
label_name = item[2] + ' ' + item[3]
img = cv2.imread(img_name)
label = np.load(label_name)
img_array = np.array(img, dtype=np.float32)
imgs.append(img_array/255.0)
label_array = np.array(label, dtype=np.float32)
lables_array_norm = (label_array)/(255.0*1.1)
labels.append(lables_array_norm)
batch.append(imgs)
batch.append(labels)
return batch
def __call__(self, batch_num):
if(self.index+batch_num) <= self.num_data:
batch_list = self.train_data_list[self.index:(self.index+batch_num)]
batch_data = self.getBatch(batch_list)
self.index += batch_num
return batch_data
elif self.index < self.num_data:
batch_list = self.train_data_list[self.index:self.num_data]
batch_data = self.getBatch(batch_list)
self.index = 0
return batch_data
else:
self.index = 0
batch_list = self.train_data_list[self.index:(self.index+batch_num)]
batch_data = self.getBatch(batch_list)
self.index += batch_num
return batch_data
'''''
def main(args):
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
batch_size = args.batch_size
epoch = args.epoch
train_data_file = args.train_data_file
learning_rate = args.learning_rate
model_path = args.model_path
#save_dir = args.model_path
#if not os.path.exists(save_dir):
# os.makedirs(save_dir)
# mask
weight_mask_path = 'mask/uv_weight_mask.png'
face_mask_path = 'mask/uv_face_mask.png'
weight_mask = cv2.imread(weight_mask_path, cv2.IMREAD_GRAYSCALE).astype('float32')
weight_mask = weight_mask/255.0
face_mask = cv2.imread(face_mask_path, cv2.IMREAD_GRAYSCALE).astype('float32')
face_mask = face_mask/255.0
final_mask = np.multiply(face_mask, weight_mask)
final_weight_mask = np.zeros(shape=(1,256,256,3)).astype('float32')
final_weight_mask[0,:,:,0]=final_mask
final_weight_mask[0,:,:,1]=final_mask
final_weight_mask[0,:,:,2]=final_mask
#print(final_weight_mask.shape)
final_weight_mask = final_weight_mask.transpose((0,3,1,2))
#print(final_weight_mask.shape)
final_weight_mask = torch.from_numpy(final_weight_mask)
final_weight_mask = final_weight_mask.to(device=torch.device('cuda'))
#print(final_weight_mask.shape)
# mask
#mask = cv2.imread('mask/weight_mask_final.jpg')
#mask = mask.astype("float32")/255.0
#mask = mask.transpose((2,0,1))
#mask = torch.from_numpy(mask)#256x256x3 tensor
#mask = mask.cuda()
#print(mask.shape)
# load training data 300wlp
data300wlp = Dataset300WLP(train_data_file,
transform=transforms.Compose([
ToTensor(),
ToNormalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])]))
dataloader300wlp = DataLoader(data300wlp, batch_size=batch_size, shuffle=True, num_workers=8, pin_memory=True)
print(data300wlp.__len__())
# network
net = ResFCN256()
#device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
device = torch.device('cuda')
net.to(device)
#print("net is:\n",net)
# loss and optimizer
optimizer = optim.Adam(net.parameters(), lr=learning_rate)
#scheduler = optim.lr_scheduler.StepLR(optimizer, 5, 0.5) # decays half after each 5 epoches
scheduler = optim.lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.5)
criterion = nn.L1Loss(reduction="mean")
print(optimizer.defaults['lr'])
# train
for ep in range(epoch):
running_loss = 0.0
for i,data in enumerate(dataloader300wlp, 0):
img, label = data['origin_img'].cuda(), data['gt_label'].cuda()
#img = img.cuda()
#label = label.cuda()
optimizer.zero_grad()
#print(img.shape)
#print(label.shape)
#print(mask.shape)
outputs = net(img)
#print(outputs.shape)
#outputs = torch.mul(outputs, mask)
#label = torch.mul(label, mask)
outputs = torch.mul(outputs, final_weight_mask)
label = torch.mul(label, final_weight_mask)
loss = criterion(outputs, label)
loss.backward()
optimizer.step()
#scheduler.step()
running_loss += loss.item()
if i%20 == 19:
#print('[%d, %5d] loss: %.3f' %
# (ep + 1, i + 1, running_loss / 20))
print('[epoch: %d, %d, lr: %.10f] loss: ' % (ep+1, i+1, optimizer.param_groups[0]['lr']), running_loss/20)
running_loss = 0.0
scheduler.step()
print("Final Loss: ", loss.item())
print("Finished Training!")
# save model
torch.save(net.state_dict(), model_path)
if __name__ == '__main__':
par = argparse.ArgumentParser(description='my3DFaceRecon')
par.add_argument('--train_data_file', default='data/300wlp_all.txt', type=str, help='Training data file (txt)')
par.add_argument('--learning_rate', default=0.0001, type=float, help='Learning rate')
par.add_argument('--epoch', default=20, type=int, help='Epoches to train')
par.add_argument('--batch_size', default=16, type=int, help='Batch size')
par.add_argument('--model_path', default='model/model1.pth', help='Model path')
par.add_argument('--gpu', default='0', type=str, help='GPU ID')
main(par.parse_args())