STDL/load_dataset.py at main · ChenxiangMA/STDL · GitHub

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
import os
import logging

from torchvision import datasets, transforms
import torch

from datasets.autoaugment import CIFAR10Policy, Cutout


def load_dataset(args=None):
    if args.dataset in ['cifar10', 'cifar100']:
        args.image_size = 32
        if args.dataset == 'cifar10':
            num_classes = 10
            normalize = transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010))
        elif args.dataset == 'cifar100':
            num_classes = 100
            normalize = transforms.Normalize((0.5071, 0.4867, 0.4408), (0.2675, 0.2565, 0.2761))
        train_augmentation = [transforms.RandomCrop(32, padding=4),
                              transforms.RandomHorizontalFlip()]
        if args.auto_aug:
            train_augmentation.append(CIFAR10Policy())
        train_augmentation.append(transforms.ToTensor())
        if args.cutout:
            train_augmentation.append(Cutout(n_holes=1, length=16))
        train_augmentation.append(normalize)
        train_transform = transforms.Compose(train_augmentation)
        test_transform = transforms.Compose([
            transforms.ToTensor(),
            normalize
        ])

        train_dataset = datasets.__dict__[args.dataset.upper()]('../datasets/' + args.dataset, download=True,
                                                                transform=train_transform, train=True
                                                                )
        train_sampler = None
        batch_size = args.batch_size
        train_loader = torch.utils.data.DataLoader(train_dataset,
                                                   batch_size=batch_size,
                                                   num_workers=args.workers,
                                                   shuffle=(train_sampler is None),
                                                   pin_memory=True,
                                                   sampler=train_sampler,
                                                   )
        val_loader = torch.utils.data.DataLoader(
            datasets.__dict__[args.dataset.upper()]('../datasets/' + args.dataset, transform=test_transform,
                                                    download=True,
                                                    train=False),
            batch_size=batch_size,
            num_workers=args.workers,
            pin_memory=True,
            shuffle=False)
    elif args.dataset == 'svhn':
        args.image_size = 32
        num_classes = 10
        normalize = transforms.Normalize((0.431, 0.430, 0.446), (0.197, 0.198, 0.199))
        train_transform = transforms.Compose([
            transforms.RandomCrop(32, padding=2),
            transforms.ToTensor(),
            normalize
        ])
        test_transform = transforms.Compose([
            transforms.ToTensor(),
            normalize
        ])

        train_dataset = datasets.__dict__[args.dataset.upper()]('../datasets/' + args.dataset, download=True,
                                                                transform=train_transform, split="train"
                                                                )

        train_sampler = None
        batch_size = args.batch_size
        train_loader = torch.utils.data.DataLoader(train_dataset,
                                                   batch_size=batch_size,
                                                   num_workers=args.workers,
                                                   shuffle=(train_sampler is None),
                                                   pin_memory=True,
                                                   sampler=train_sampler,
                                                   )
        val_loader = torch.utils.data.DataLoader(
            datasets.__dict__[args.dataset.upper()]('../datasets/' + args.dataset, transform=test_transform,
                                                    download=True,
                                                    split="test"),
            batch_size=batch_size,
            num_workers=args.workers,
            pin_memory=True,
            shuffle=False)
    elif args.dataset == 'imagenet':
        num_classes = 1000
        traindir = os.path.join(args.data, 'train')
        valdir = os.path.join(args.data, 'val')
        normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                         std=[0.229, 0.224, 0.225])

        train_transform = transforms.Compose([
            transforms.RandomResizedCrop(224),
            transforms.RandomHorizontalFlip(),
            transforms.ToTensor(),
            normalize,
        ])
        train_dataset = datasets.ImageFolder(
            traindir,
            train_transform
        )
        args.image_size = 224

        train_sampler = None
        batch_size = args.batch_size

        train_loader = torch.utils.data.DataLoader(
            train_dataset, batch_size=batch_size, shuffle=(train_sampler is None),
            num_workers=args.workers, pin_memory=True, sampler=train_sampler)

        val_loader = torch.utils.data.DataLoader(
            datasets.ImageFolder(valdir, transforms.Compose([
                transforms.Resize(256),
                transforms.CenterCrop(224),
                transforms.ToTensor(),
                normalize,
            ])),
            batch_size=batch_size, shuffle=False,
            num_workers=args.workers, pin_memory=True)
    elif args.dataset == 'cifar10dvs':
        from datasets.cifardvs.cifar10_dvs import CIFAR10DVS
        from datasets.cifardvs.augmentation import ToPILImage, Resize, ToTensor
        num_classes = 10

        transform_train = transforms.Compose([
            ToPILImage(),
            # Resize(48),
            ToTensor(),
        ])
        logging.info("no augmentation")
        transform_test = transforms.Compose([
            ToPILImage(),
            # Resize(48),
            ToTensor(),
        ])
        args.image_size = 128
        train_dataset = CIFAR10DVS('../datasets/cifar10dvs', train=True, use_frame=True, frames_num=args.time_window, split_by='number',
                              normalization=None, transform=transform_train)
        test_dataset = CIFAR10DVS('../datasets/cifar10dvs', train=False, use_frame=True, frames_num=args.time_window, split_by='number',
                             normalization=None, transform=transform_test)
        train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=args.batch_size, shuffle=True, num_workers=args.workers, pin_memory=False)
        val_loader = torch.utils.data.DataLoader(test_dataset, batch_size=args.batch_size, shuffle=False, num_workers=args.workers, pin_memory=False)

    elif args.dataset == 'hardvs':
        from datasets.hardvs.dvs_har import HARDVS
        from datasets.cifardvs.augmentation import RandomResizedCrop, ToPILImage, HorizontalFlip, ToTensor, Resize, CenterCrop, ToPILImageTensor

        transform_train = transforms.Compose(
            [
                ToPILImageTensor(),
                RandomResizedCrop(),
                HorizontalFlip(),
                ToTensor(),
            ]
        )
        transform_test = transforms.Compose(
            [ToPILImageTensor(), Resize(256), CenterCrop(224), ToTensor()]
        )
        # data_path = '/home/chenxiangma/projects/datasets/hardvs'
        data_path = '/datasets/hardvs'
        train_dataset = HARDVS(root=data_path, train=True, data_type='frame', frames_number=args.time_window, split_by='number', transform=transform_train)
        train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=args.batch_size, shuffle=True, num_workers=args.workers,
                                            drop_last=True,
                                            pin_memory=True)

        test_dataset = HARDVS(root=data_path, train=False, data_type='frame', frames_number=args.time_window, split_by='number', transform=transform_test)
        val_loader = torch.utils.data.DataLoader(test_dataset, batch_size=args.batch_size, shuffle=False, num_workers=args.workers,
                                           drop_last=False,
                                           pin_memory=True)
        args.image_size = 112
        num_classes = 300
    else:
        raise Exception('No valid dataset is specified.')
    return train_loader, val_loader, num_classes