/
datasets.py
188 lines (164 loc) · 8.75 KB
/
datasets.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
186
187
188
import glob
import json
import os
import os.path as osp
import torch
from torchvision import datasets
from diffusion.utils import DATASET_ROOT, get_classes_templates
from diffusion.dataset.objectnet import ObjectNetBase
from diffusion.dataset.imagenet_classnames import get_classnames
from imagenetv2_pytorch import ImageNetV2Dataset
from PIL import Image
IMAGENET_A_CLASSES = [
6, 11, 13, 15, 17, 22, 23, 27, 30, 37, 39, 42, 47, 50, 57, 70, 71, 76, 79, 89, 90, 94, 96, 97, 99, 105, 107, 108,
110, 113, 124, 125, 130, 132, 143, 144, 150, 151, 207, 234, 235, 254, 277, 283, 287, 291, 295, 298, 301, 306, 307,
308, 309, 310, 311, 313, 314, 315, 317, 319, 323, 324, 326, 327, 330, 334, 335, 336, 347, 361, 363, 372, 378, 386,
397, 400, 401, 402, 404, 407, 411, 416, 417, 420, 425, 428, 430, 437, 438, 445, 456, 457, 461, 462, 470, 472, 483,
486, 488, 492, 496, 514, 516, 528, 530, 539, 542, 543, 549, 552, 557, 561, 562, 569, 572, 573, 575, 579, 589, 606,
607, 609, 614, 626, 627, 640, 641, 642, 643, 658, 668, 677, 682, 684, 687, 701, 704, 719, 736, 746, 749, 752, 758,
763, 765, 768, 773, 774, 776, 779, 780, 786, 792, 797, 802, 803, 804, 813, 815, 820, 823, 831, 833, 835, 839, 845,
847, 850, 859, 862, 870, 879, 880, 888, 890, 897, 900, 907, 913, 924, 932, 933, 934, 937, 943, 945, 947, 951, 954,
956, 957, 959, 971, 972, 980, 981, 984, 986, 987, 988
]
OBJECTNET_CLASSES = [
409, 412, 414, 418, 419, 423, 434, 440, 444, 446, 455, 457, 462, 463, 470, 473, 479, 487, 499, 504, 507, 508, 518,
530, 531, 533, 539, 543, 545, 549, 550, 559, 560, 563, 567, 578, 587, 588, 589, 601, 606, 608, 610, 618, 619, 620,
623, 626, 629, 630, 632, 644, 647, 651, 655, 658, 659, 664, 671, 673, 677, 679, 689, 694, 695, 696, 700, 703, 720,
721, 725, 728, 729, 731, 732, 737, 740, 742, 749, 752, 759, 761, 765, 769, 770, 772, 773, 774, 778, 783, 790, 792,
797, 804, 806, 809, 811, 813, 828, 834, 837, 841, 842, 846, 849, 850, 851, 859, 868, 879, 882, 883, 893, 898, 902,
906, 907, 909, 923, 930, 950, 951, 954, 968, 999
]
with open('diffusion/imagenet_class_index.json', 'r') as f:
IMAGENET_CLASS_INDEX = json.load(f)
FOLDER_TO_CLASS = {folder: int(i) for i, (folder, _) in IMAGENET_CLASS_INDEX.items()}
class MNIST(datasets.MNIST):
"""Simple subclass to override the property"""
class_to_idx = {str(i): i for i in range(10)}
class ImageNetV2(ImageNetV2Dataset):
def __init__(
self,
variant="matched-frequency",
root=DATASET_ROOT,
transform=None,
):
super(ImageNetV2, self).__init__(variant=variant,
location=root,
transform=transform)
self.fnames.sort()
class ImageNetA(torch.utils.data.Dataset):
def __init__(
self,
root=DATASET_ROOT,
transform=lambda x: x,
):
super(ImageNetA, self).__init__()
self.root = os.path.join(root, 'imagenet-a')
self._images = glob.glob(self.root + '/*/*')
self._images.sort()
self.transform = transform
assert len(self._images) == 7500, f"Expected {len(self)} images, found {len(self._images)}"
def __len__(self):
return 7500
def path_to_cls(self, path):
folder = path.split('/')[-2]
return FOLDER_TO_CLASS[folder]
def __getitem__(self, index):
filepath = self._images[index]
img, label = Image.open(filepath).convert("RGB"), self.path_to_cls(filepath)
if self.transform is not None:
img = self.transform(img)
return img, label
def get_target_dataset(name: str, train=False, transform=None, target_transform=None):
"""Get the torchvision dataset that we want to use.
If the dataset doesn't have a class_to_idx attribute, we add it.
Also add a file-to-class map for evaluation
"""
if name == "cifar10":
dataset = datasets.CIFAR10(root=DATASET_ROOT, train=train, transform=transform,
target_transform=target_transform, download=True)
elif name == "stl10":
dataset = datasets.STL10(root=DATASET_ROOT, split="train" if train else "test", transform=transform,
target_transform=target_transform, download=True)
dataset.class_to_idx = {cls: i for i, cls in enumerate(dataset.classes)}
elif name == "pets":
dataset = datasets.OxfordIIITPet(root=DATASET_ROOT, split="trainval" if train else "test", transform=transform,
target_transform=target_transform, download=True)
# lower case every key in the class_to_idx
dataset.class_to_idx = {k.lower(): v for k, v in dataset.class_to_idx.items()}
dataset.file_to_class = {f.name.split('.')[0]: l for f, l in zip(dataset._images, dataset._labels)}
elif name == "flowers":
dataset = datasets.Flowers102(root=DATASET_ROOT, split="train" if train else "test", transform=transform,
target_transform=target_transform, download=True)
classes = list(get_classes_templates('flowers')[0].keys()) # in correct order
dataset.class_to_idx = {cls: i for i, cls in enumerate(classes)}
dataset.file_to_class = {f.name.split('.')[0]: l for f, l in zip(dataset._image_files, dataset._labels)}
elif name == "aircraft":
dataset = datasets.FGVCAircraft(root=DATASET_ROOT, split="trainval" if train else "test", transform=transform,
target_transform=target_transform, download=True)
# replace backslash with underscore -> need to be dirs
dataset.class_to_idx = {
k.replace('/', '_'): v
for k, v in dataset.class_to_idx.items()
}
dataset.file_to_class = {
fn.split("/")[-1].split(".")[0]: lab
for fn, lab in zip(dataset._image_files, dataset._labels)
}
# dataset.file_to_class = {
# fn.split("/")[-1].split(".")[0]: lab
# for fn, lab in zip(dataset._image_files, dataset._labels)
# }
elif name == "food":
dataset = datasets.Food101(root=DATASET_ROOT, split="train" if train else "test", transform=transform,
target_transform=target_transform, download=True)
dataset.file_to_class = {
f.name.split(".")[0]: dataset.class_to_idx[f.parents[0].name]
for f in dataset._image_files
}
elif name == "eurosat":
if train:
raise ValueError("EuroSAT does not have a train split.")
dataset = datasets.EuroSAT(root=DATASET_ROOT, transform=transform, target_transform=target_transform,
download=True)
elif name == 'imagenet':
assert not train
dataset = datasets.ImageFolder(root=osp.join(DATASET_ROOT, 'imagenet/val'),
transform=transform,
target_transform=target_transform)
dataset.class_to_idx = None
dataset.classes = get_classnames('openai')
dataset.file_to_class = None
elif name == 'objectnet':
base = ObjectNetBase(transform, DATASET_ROOT)
dataset = base.get_test_dataset()
dataset.class_to_idx = dataset.label_map
dataset.file_to_class = None # todo
elif name == 'imagenetv2':
assert not train
dataset = ImageNetV2(root=DATASET_ROOT, transform=transform)
dataset.file_to_class = None
dataset.class_to_idx = None
elif name == 'imagenetA':
dataset = ImageNetA(root=DATASET_ROOT, transform=transform)
dataset.file_to_class = None
dataset.class_to_idx = None
elif name == "caltech101":
if train:
raise ValueError("Caltech101 does not have a train split.")
dataset = datasets.Caltech101(root=DATASET_ROOT, target_type="category", transform=transform,
target_transform=target_transform, download=True)
dataset.class_to_idx = {cls: i for i, cls in enumerate(dataset.categories)}
dataset.file_to_class = {str(idx): dataset.y[idx] for idx in range(len(dataset))}
elif name == "mnist":
dataset = MNIST(root=DATASET_ROOT, train=train, transform=transform, target_transform=target_transform,
download=True)
else:
raise ValueError(f"Dataset {name} not supported.")
if name in {'mnist', 'cifar10', 'stl10', 'aircraft'}:
dataset.file_to_class = {
str(idx): dataset[idx][1]
for idx in range(len(dataset))
}
assert hasattr(dataset, "class_to_idx"), f"Dataset {name} does not have a class_to_idx attribute."
assert hasattr(dataset, "file_to_class"), f"Dataset {name} does not have a file_to_class attribute."
return dataset