-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
166 lines (146 loc) · 8.23 KB
/
utils.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
import torch
import torchvision
import os
import numpy as np
import datasets
import torch.nn.functional as F
import copy
from PIL import Image
from mypath import Path
def multi_class_loss(pred, target):
pred = F.log_softmax(pred, dim=1)
loss = - torch.sum(target*pred, dim=1)
return loss
def mixup_data(x, y, alpha=1.0, device='cuda'):
'''Returns mixed inputs, pairs of targets, and lambda'''
if alpha > 0:
lam = np.random.beta(alpha, alpha)
lam = max(lam, 1-lam)
else:
lam = 1
device = x.get_device()
batch_size = x.size()[0]
index = torch.randperm(batch_size).to(device)
mixed_x = lam * x + (1 - lam) * x[index, :]
y_a, y_b = y, y[index]
return mixed_x, y_a, y_b, lam, index
def make_data_loader(args, **kwargs):
if args.dataset == 'cifar10':
mean = [0.4914, 0.4822, 0.4465]
std = [0.2023, 0.1994, 0.2010]
size1 = 32
size = 32
elif args.dataset == 'cifar100':
mean = [0.5071, 0.4867, 0.4408]
std = [0.2675, 0.2565, 0.2761]
size1 = 32
size = 32
elif args.dataset == "miniimagenet_preset":
mean = [0.4728, 0.4487, 0.4031]
std = [0.2744, 0.2663 , 0.2806]
if args.net!="inception":
size1 = 32
size = 32
else:
size1 = 84
size = 84
elif args.dataset == 'webvision':
mean = [0.485, 0.456, 0.406]
std = [0.229, 0.224, 0.225]
if args.sup:#Supervised training
size1 = 256
size = 227
else:
size1 = 84
size = 84
transform_train = torchvision.transforms.Compose([
torchvision.transforms.Resize(size1, interpolation=Image.BICUBIC),
torchvision.transforms.RandomCrop(size, padding=4),
torchvision.transforms.RandomHorizontalFlip(),
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize(mean, std),
])
transform_unsup = torchvision.transforms.Compose([
torchvision.transforms.Resize(size1, interpolation=Image.BICUBIC),
torchvision.transforms.RandomResizedCrop(size, interpolation=Image.BICUBIC),
torchvision.transforms.RandomHorizontalFlip(p=0.5),
torchvision.transforms.RandomApply([torchvision.transforms.ColorJitter(0.4, 0.4, 0.4, 0.1)], p=0.8),
torchvision.transforms.RandomGrayscale(p=0.2),
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize(mean, std),
])
transform_test = torchvision.transforms.Compose([
torchvision.transforms.Resize(size1, interpolation=Image.BICUBIC),
torchvision.transforms.CenterCrop(size),
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize(mean, std)
])
if args.dataset == "cifar10":
from datasets.cifar import CIFAR10
trainset = CIFAR10(Path.db_root_dir("cifar10"), ood_noise=args.ood_noise, id_noise=args.id_noise, train=True, transform=transform_train, cont=args.cont, transform_unsup=transform_unsup, consistency=len(args.id_in) > 0, corruption=args.corruption)
testset = CIFAR10(Path.db_root_dir("cifar10"), ood_noise=args.ood_noise, id_noise=args.id_noise, classes_id=trainset.classes_id, train=False, transform=transform_test)
elif args.dataset == "cifar100":
from datasets.cifar import CIFAR100
trainset = CIFAR100(Path.db_root_dir("cifar100"), ood_noise=args.ood_noise, id_noise=args.id_noise, train=True, transform=transform_train, cont=args.cont, transform_unsup=transform_unsup, consistency=len(args.id_in) > 0, corruption=args.corruption)
testset = CIFAR100(Path.db_root_dir("cifar100"), ood_noise=args.ood_noise, id_noise=args.id_noise, classes_id=trainset.classes_id, train=False, transform=transform_test)
elif args.dataset == "miniimagenet_preset":
from datasets.miniimagenet_preset import make_dataset, MiniImagenet84
train_data, train_labels, val_data, val_labels, test_data, test_labels, clean_noisy = make_dataset(noise_ratio=args.noise_ratio)
trainset = MiniImagenet84(train_data, train_labels, transform=transform_train, cont=args.cont, transform_unsup=transform_unsup, consistency=len(args.id_in) > 0)
testset = MiniImagenet84(val_data, val_labels, transform=transform_test)
elif args.dataset == "webvision":
from datasets.webvision import webvision_dataset
trainset = webvision_dataset(transform=transform_train, mode="train", num_classes=50, cont=args.cont, transform_unsup=transform_unsup, consistency=len(args.id_in) > 0)
testset = webvision_dataset(transform=transform_test, mode="test", num_classes=50)
else:
raise NotImplementedError("Dataset {} is not implemented".format(args.dataset))
trackset = copy.deepcopy(trainset)
trackset.cont = False
trackset.transform = transform_test
warmupset = copy.deepcopy(trainset)
warmupset.cont = False
warmupset.consistency = False
track_loader = torch.utils.data.DataLoader(trackset, batch_size=args.batch_size, shuffle=True, **kwargs)
test_loader = torch.utils.data.DataLoader(testset, batch_size=256, shuffle=False, **kwargs)
if args.sup:
if args.ts: #Equal sampler
if len(args.id_ood) == 0 and len(args.id_in) > 0:
from ThreeSampler import TwoStreamBatchSampler
sampler = TwoStreamBatchSampler(args.id_clean, args.id_in, batch_size=args.batch_size)
train_loader = torch.utils.data.DataLoader(trainset, batch_sampler=sampler, **kwargs)
sampler = torch.utils.data.SubsetRandomSampler(args.id_clean)
warmup_loader = torch.utils.data.DataLoader(warmupset, batch_size=args.batch_size, sampler=sampler, **kwargs)
elif len(args.id_in) > 0:
from ThreeSampler import ThreeStreamBatchSampler
sampler = ThreeStreamBatchSampler(args.id_clean, args.id_in, args.id_ood, batch_size=args.batch_size)
train_loader = torch.utils.data.DataLoader(trainset, batch_sampler=sampler, **kwargs)
sampler = torch.utils.data.SubsetRandomSampler(args.id_clean)
warmup_loader = torch.utils.data.DataLoader(warmupset, batch_size=args.batch_size, sampler=sampler, **kwargs)
else:
from ThreeSampler import TwoStreamBatchSampler
sampler = TwoStreamBatchSampler(args.id_clean, args.id_ood, batch_size=args.batch_size)
train_loader = torch.utils.data.DataLoader(trainset, batch_sampler=sampler, **kwargs)
sampler = torch.utils.data.SubsetRandomSampler(args.id_clean)
warmup_loader = torch.utils.data.DataLoader(warmupset, batch_size=args.batch_size, sampler=sampler, **kwargs)
else:
train_loader = torch.utils.data.DataLoader(trainset, batch_size=args.batch_size, shuffle=True, **kwargs)
sampler = torch.utils.data.SubsetRandomSampler(args.id_clean)
warmup_loader = torch.utils.data.DataLoader(warmupset, batch_size=args.batch_size, sampler=sampler, **kwargs)
return warmup_loader, train_loader, test_loader, track_loader
elif args.dataset != 'clothing':
train_loader = torch.utils.data.DataLoader(trainset, batch_size=args.batch_size, shuffle=True, **kwargs)
else:
indexs = torch.randperm(len(trainset))[:args.batch_size*1000]
sampler = torch.utils.data.SubsetRandomSampler(indexs)
train_loader = torch.utils.data.DataLoader(trainset, batch_size=args.batch_size, sampler=sampler, **kwargs)
return train_loader, test_loader, track_loader
def create_save_folder(args):
if not os.path.isdir(args.save_dir):
os.mkdir(args.save_dir)
if not os.path.isdir(os.path.join(args.save_dir, args.net + '_' + args.dataset)):
os.mkdir(os.path.join(args.save_dir, args.net + '_' + args.dataset))
if not os.path.isdir(os.path.join(args.save_dir, args.net + '_' + args.dataset, args.exp_name)):
os.mkdir(os.path.join(args.save_dir, args.net + '_' + args.dataset, args.exp_name))
if not os.path.isdir(os.path.join(args.save_dir, args.net + '_' + args.dataset, args.exp_name, str(args.seed))):
os.mkdir(os.path.join(args.save_dir, args.net + '_' + args.dataset, args.exp_name, str(args.seed)))
return