-
Notifications
You must be signed in to change notification settings - Fork 0
/
Multi_classification.py
144 lines (129 loc) · 6.3 KB
/
Multi_classification.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
import os
import matplotlib.pyplot as plt
from torch.utils.data import DataLoader
import torch
from torchvision.transforms import transforms
from torch.utils.data import Dataset
from torchvision.transforms import functional as F
from PIL import Image
import pandas as pd
import random
from torch import optim
from torch.optim import lr_scheduler
import copy
from Multi_NetWork import Net
import datetime
ROOT_DIR = '../Dataset/'
TRAIN_DIR = 'train/'
VAL_DIR = 'val/'
TRAIN_ANNO = './csv/Multi_train_make_anno_good.csv'
VAL_ANNO = './csv/Multi_val_make_anno_good.csv'
CLASSES = ['Mammals', 'Birds']
SPECIES = ['rabbits', 'rats', 'chickens']
class MyDataSet(Dataset):
def __init__(self, root_dir, anno_file, transform=False):
self.root_dir = root_dir
self.anno_file = anno_file
self.transform = transform
if (not os.path.isfile(anno_file)):
print(self.anno_file, ' is not exist!')
self.file_info = pd.read_csv(self.anno_file, index_col=0)
def __len__(self):
return len(self.file_info)
def __getitem__(self, idx):
image_path = self.file_info['path'][idx]
if (not os.path.isfile(image_path)):
print(image_path, ' is not exist!')
image = Image.open(image_path).convert('RGB')
label_classes = self.file_info['classes'][idx]
label_species = self.file_info['species'][idx]
sample = {'image': image, 'classes': label_classes, 'species': label_species}
if (self.transform):
sample['image'] = self.transform(image)
return sample
train_transform = transforms.Compose([transforms.Resize((500, 500)),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])])
test_transform = transforms.Compose([transforms.Resize((500, 500)),
transforms.ToTensor(),
transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])])
train_dataset = MyDataSet(root_dir=ROOT_DIR + TRAIN_DIR, anno_file=TRAIN_ANNO, transform=train_transform)
test_dataset = MyDataSet(root_dir=ROOT_DIR + VAL_DIR, anno_file=VAL_ANNO, transform=test_transform)
train_loader = DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)
test_loader = DataLoader(dataset=test_dataset, shuffle=True)
data_loader = {'train': train_loader, 'val': test_loader}
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
print('device:', device)
def visual_dataset():
print(len(train_dataset))
idx = random.randint(0, len(train_dataset))
sample = train_loader.dataset[idx]
print(idx, sample['image'].shape, CLASSES[sample['classes']],SPECIES[sample['species']])
plt.imshow(transforms.ToPILImage()(sample['image']))
plt.show()
# visual_dataset()
def train_model(net, optimizer, criterion, scheduler, epochs=50,rate=0.5):
Loss_list = {'train': [], 'val': []}
Accuracy_list = {'train': [], 'val': []}
best_model = copy.deepcopy(net.state_dict())
best_acc = 0.0
for epoch in range(epochs):
print('Epoch:{}/{}'.format(epoch, epochs - 1))
print('lr', optimizer.state_dict()['param_groups'][0]['lr'])
for phase in ['train', 'val']:
if (phase == 'train'):
net.train()
else:
net.eval()
running_loss = .0
corrects_classes = 0
start_time=datetime.datetime.now()
for idx, data in enumerate(data_loader[phase]):
inputs = data['image'].to(device)
labels_classes = data['classes'].to(device)
labels_species = data['species'].to(device)
optimizer.zero_grad()
with torch.set_grad_enabled(phase == 'train'):
outputs_class, outputs_species = net(inputs)
_, pre_classes = torch.max(outputs_class, 1)
_, pre_species = torch.max(outputs_species, 1)
loss_class = criterion(outputs_class, labels_classes)
loss_species = criterion(outputs_species, labels_species)
loss = rate*loss_class + (1-rate)*loss_species
if (phase == 'train'):
loss.backward()
optimizer.step()
running_loss += loss.item() * inputs.size(0)
corrects_classes += torch.sum((pre_classes == labels_classes) & (pre_species == labels_species))
epoch_loss = running_loss / len(data_loader[phase].dataset)
epoch_acc = 1.0 * corrects_classes.double() / len(data_loader[phase].dataset)
print('{} loss: {:.3f} acc_rate: {:.3f} runtime(s): {}'.format(phase, epoch_loss, epoch_acc,
datetime.datetime.now()-start_time))
Loss_list[phase].append(epoch_loss)
Accuracy_list[phase].append(epoch_acc)
if (phase == 'val' and epoch_acc > best_acc):
best_acc = epoch_acc
best_model = copy.deepcopy(net.state_dict())
print('Best val classes Acc: {:.4f}'.format(best_acc))
scheduler.step()
net.load_state_dict(best_model)
time_str=datetime.datetime.now().strftime('%Y%m%d%H%M%S')
torch.save(net.state_dict(), './model_pt/best_model_%s.pt'%time_str)
print('Final val Classes Acc: {:.4f}'.format(best_acc))
return net, Loss_list, Accuracy_list
net = Net().to(device)
optimizer = optim.SGD(net.parameters(), lr=.01, momentum=0.9)
# optimizer=optim.Adam(net.parameters(),lr=0.01)
criterion = torch.nn.CrossEntropyLoss()
epoch_lr_scheduler = lr_scheduler.StepLR(optimizer, step_size=30, gamma=0.7)
net,Loss_list,Accuracy_list=train_model(net, optimizer, criterion, epoch_lr_scheduler, epochs=100,rate=0.8)
#save Loss_list
anno_loss_list=pd.DataFrame(Loss_list)
time_str=datetime.datetime.now().strftime('%Y%m%d%H%M%S')
anno_loss_list.to_csv('./csv/loss_list_%s.csv' % time_str)
#save Accuracy_list
Accuracy_list={'train':torch.tensor(Accuracy_list['train']).numpy(),'val':torch.tensor(Accuracy_list['val']).numpy()}
anno_acc_list=pd.DataFrame(Accuracy_list)
time_str=datetime.datetime.now().strftime('%Y%m%d%H%M%S')
anno_acc_list.to_csv('./csv/acc_list_%s.csv' % time_str)