-
Notifications
You must be signed in to change notification settings - Fork 0
/
AutoEncoder.py
175 lines (146 loc) · 5.95 KB
/
AutoEncoder.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
import pdb
from dataset import *
from utils import save_csv, clearDir, calculate_time
from torch.utils.data import DataLoader
from model import AutoEncoderConv, AutoEncoderConv_Lite
from torch import nn
from torch.optim import lr_scheduler
import matplotlib.pyplot as plt
import torchvision.transforms as transforms
from torchvision.datasets import ImageFolder
from torch.cuda.amp import autocast, GradScaler
import argparse
@calculate_time
def train(model, data_dir, save_model_path='checkpoints/AutoEncoder/', log_path='log/AutoEncoder/'):
clearDir(save_model_path)
clearDir(log_path)
train_loss_log = log_path + 'AutoEncoder_train_loss.csv'
epochs = 50
batch_size = 128
learning_rate = 0.001
num_workers = torch.cuda.device_count()*4
# Initialize the autoencoder
model.train()
# Define transform
transform = transforms.Compose([
transforms.Resize((64, 64)),
# transforms.RandomHorizontalFlip(p=0.5),
transforms.RandomRotation([-30,30], interpolation=transforms.InterpolationMode.BILINEAR, expand=False),
transforms.ToTensor(),
])
# Load dataset
dataset = ImageFolder(root=data_dir, transform=transform)
# Define the dataloader
dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True, num_workers=num_workers)
# Move the model to GPU
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
torch.backends.cudnn.benchmark = True if device==torch.device('cuda') else False
print("Now use device: " + str(device))
model.to(device)
# Define the loss function and optimizer
loss_func = nn.MSELoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
scheduler = lr_scheduler.StepLR(optimizer, 20, gamma=0.1)
min_loss = None
scaler = GradScaler()
# Train the autoencoder
for epoch in range(epochs):
for batch_data in dataloader:
batch_data, _ = batch_data
batch_data = batch_data.to(device)
optimizer.zero_grad()
# ===================forward=====================
# Runs the forward pass under autocast.
with autocast():
output = model(batch_data)
loss = loss_func(output, batch_data)
# ===================backward====================
scaler.scale(loss).backward()
scaler.step(optimizer)
scaler.update()
scheduler.step()
# ===================log========================
loss_log = loss.cpu().detach().numpy()
save_csv(loss_log, train_loss_log)
if (epoch + 1) % 5== 0:
print('epoch [{}/{}], loss:{:.4f}'.format(epoch + 1, epochs, loss.item()))
if (epoch + 1) % 10 == 0:
torch.save(model.state_dict(), save_model_path + 'model_epoch_{}.pt'.format(epoch + 1))
if (min_loss!=None and min_loss > loss):
torch.save(model.state_dict(), save_model_path + 'model_epoch_final.pt')
min_loss = loss
elif(not min_loss):
min_loss = loss
@calculate_time
def test(model, test_dir='', model_path='checkpoints/AutoEncoder/model_epoch_final.pt'):
batch_size = 1
num_workers = torch.cuda.device_count()*4
# Move the model to GPU
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print("Now use device: " + str(device))
model.to(device)
model.load_state_dict(torch.load(model_path))
model.eval()
# Define transform
transform = transforms.Compose([
transforms.Resize((64, 64)),
transforms.ToTensor(),
])
# Load dataset
dataset = ImageFolder(root=test_dir, transform=transform)
# Define the dataloader
test_dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=False, num_workers=num_workers)
loss_func = nn.MSELoss()
with torch.no_grad():
for batch_data in test_dataloader:
batch_data, _ = batch_data
batch_data = batch_data.to(device)
output = model(batch_data)
loss = loss_func(output, batch_data)
print('Output Score: {}'.format(round(float(loss), 3)))
@calculate_time
def vis(model, test_dir='', model_path='checkpoints/AutoEncoder/model_epoch_final.pt'):
batch_size = 128
num_workers = torch.cuda.device_count()*4
loss_func = nn.MSELoss()
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(device)
model.to(device)
model.load_state_dict(torch.load(model_path))
model.eval()
# Define transform
transform = transforms.Compose([
transforms.Resize((64, 64)),
transforms.ToTensor(),
])
dataset = ImageFolder(root=test_dir, transform=transform)
test_dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=False, num_workers=num_workers)
with torch.no_grad():
for data, _ in test_dataloader:
data = data.to(device)
recon = model(data)
break
_, ax = plt.subplots(2, 7, figsize=(15, 4))
for i in range(7):
loss = loss_func(data[i], recon[i])
loss = loss.cpu().detach().numpy()
ax[0, i].imshow(data[i].cpu().numpy().transpose((1, 2, 0)))
ax[1, i].imshow(recon[i].cpu().numpy().transpose((1, 2, 0)))
ax[0, i].axis('OFF')
ax[1, i].axis('OFF')
ax[1, i].text(0.5, -0.1, 'loss:{:.3f}'.format(loss.item()), ha='center', va='center', transform=ax[1, i].transAxes, fontsize=12)
plt.show()
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument("mode", help="choose a mode to run this Python file.")
args = parser.parse_args()
model = AutoEncoderConv()
# model = AutoEncoderConv_Lite()
data_dir = 'your_train_dateset'
test_dir = 'your_test_dateset'
if(args.mode == 'train'):
train(model, data_dir)
elif(args.mode == 'test'):
test(model, test_dir)
elif(args.mode == 'vis'):
vis(model, test_dir)