forked from lisadunlap/LADS
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdata_helpers.py
176 lines (166 loc) · 8.58 KB
/
data_helpers.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
import os, io
from tqdm import tqdm
import torch as ch
import torchvision
from torchvision import transforms
import numpy as np
# # from robustness import datasets
import PIL
from PIL import Image
import matplotlib.pyplot as plt
from omegaconf import OmegaConf
import clip
from datasets.waterbirds import Waterbirds, WaterbirdsBoring, WaterbirdsSimple, WaterbirdsOrig
from datasets.colored_mnist import ColoredMNIST, ColoredMNISTSimplified, MNIST, SVHN
from datasets.domain_net import DomainNet, DOMAINNET_CLASSES, MINI_DOMAINNET_CLASSES, MINI_DOMAINS, OneDomain, MultiDomain, DomainNetMiniAug
from datasets.cub import Cub2011Painting, Cub2011, CUB_DOMAINS, CUB_CLASSES
from datasets.office_home import OfficeHome, OFFICE_HOME_CLASSES, OFFICE_HOME_DOMAINS
from datasets.grozi import Products, GROZI_CLASSES, GROZI_DOMAINS
def get_config(name="Waterbirds"):
base_cfg = OmegaConf.load('data_configs/base.yaml')
if name == "Waterbirds":
cfg = OmegaConf.load('data_configs/waterbirds.yaml')
elif "ColoredMNIST" in name:
cfg = OmegaConf.load('data_configs/colored_mnist.yaml')
elif "DomainNet" in name:
cfg = OmegaConf.load('data_configs/domain_net.yaml')
else:
raise ValueError(f"{name} Dataset config not found")
args = OmegaConf.merge(base_cfg, cfg)
return args
def get_transform(dataset_name="Imagenet", model=None):
""""
Gets the transform for a given dataset
"""
if model in ['RN50', 'ViT-B/32']: # if we are evaluating a clip model we use its transforms
print("...loading CLIP model")
net, transform = clip.load(model)
train_transform = transform
elif dataset_name == "ColoredMNIST":
train_transform = transforms.Compose([
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize([0.4914,0.4822,0.4465], [0.2023,0.1994,0.2010])
])
transform = transforms.Compose([
transforms.Resize((224,224)),
transforms.ToTensor(),
transforms.Normalize([0.4914,0.4822,0.4465], [0.2023,0.1994,0.2010])
])
else:
train_transform = transforms.Compose([
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
transform = transforms.Compose([
transforms.Resize((224,224)),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
return train_transform, transform
def get_dataset(dataset_name, root, transform, val_transform=None):
if val_transform == None:
val_transform = transform
if dataset_name == 'Waterbirds':
args = get_config('Waterbirds')
trainset = WaterbirdsOrig(root, args, transform=transform)
valset = WaterbirdsOrig(root, args, split='val', transform=val_transform)
testset = WaterbirdsOrig(root, args, split='test', transform=val_transform)
elif dataset_name == 'Waterbirds95':
args = get_config('Waterbirds95')
trainset = WaterbirdsOrig(root, args, transform=transform)
valset = WaterbirdsOrig(root, args, split='val', transform=val_transform)
testset = WaterbirdsOrig(root, args, split='test', transform=val_transform)
elif dataset_name == 'ColoredMNISTBinary':
args = get_config('ColoredMNIST')
args.DATA.CONFOUNDING = 1.0
args.DATA.BIAS_TYPE = 'bin_blue'
trainset = ColoredMNISTSimplified('./data', args, transform=transform)
valset = ColoredMNISTSimplified('./data', args, split='val', transform=val_transform)
testset = ColoredMNISTSimplified('./data', args, split='test', transform=val_transform)
elif dataset_name == 'MNIST':
args = None
trainset = MNIST('./data', args, transform=transform)
valset = MNIST('./data', args, split='val', transform=val_transform)
testset = MNIST('./data', args, split='test', transform=val_transform)
elif dataset_name == 'SVHN':
args = None
trainset = SVHN('./data', args, transform=transform)
valset = SVHN('./data', args, split='val', transform=val_transform)
testset = SVHN('./data', args, split='test', transform=val_transform)
elif dataset_name == 'MNIST_SVHN':
args = None
trainset = MNIST('./data', args, transform=transform)
valset = MNIST('./data', args, split='test', transform=val_transform)
testset = SVHN('./data', args, split='test', transform=val_transform)
elif dataset_name == "DomainNet":
cfg = get_config('DomainNet')
trainset = DomainNet(root, cfg, split='train', transform=transform)
valset = DomainNet(root, cfg, split='val', transform=val_transform)
testset = DomainNet(root, cfg, split='test', transform=val_transform)
elif dataset_name == "DomainNetMini":
trainset = OneDomain(root, domain='sketch', split='train', transform=transform)
valset = OneDomain(root, domain='sketch', split='val', transform=val_transform)
testset = MultiDomain(root, domains=['clipart', 'painting', 'real'], split='test', transform=val_transform)
elif dataset_name == "DomainNetMiniReal":
trainset = OneDomain(root, domain='real', split='train', transform=transform)
valset = OneDomain(root, domain='real', split='val', transform=val_transform)
testset = MultiDomain(root, domains=['clipart', 'painting', 'sketch'], split='test', transform=val_transform)
elif dataset_name == "DomainNetMiniOracle":
trainset = MultiDomain(root, domains=['sketch', 'clipart', 'painting', 'real'], split='train', transform=transform)
valset = MultiDomain(root, domains=['sketch', 'clipart', 'painting', 'real'], split='val', transform=val_transform)
testset = MultiDomain(root, domains=['sketch', 'clipart', 'painting', 'real'], split='test', transform=val_transform)
elif dataset_name == "CUB":
trainset = Cub2011(root, train=True, transform=transform)
valset = Cub2011(root, train=False, transform=val_transform)
testset = Cub2011Painting('/shared/lisabdunlap/data/CUB-200-Painting', transform=val_transform)
elif dataset_name == "OfficeHomeProduct":
trainset = OfficeHome(root, domains=["Product"], train=True, transform=transform)
valset = OfficeHome(root, domains=["Product"], train=False, transform=transform)
testset = OfficeHome(root, domains=["Art", "Clipart", "Real World"],train=True, transform=transform)
elif dataset_name == "OfficeHomeClipart":
trainset = OfficeHome(root, domains=["Clipart"], train=True, transform=transform)
valset = OfficeHome(root, domains=["Clipart"], train=False, transform=transform)
testset = OfficeHome(root, domains=["Product", "Art", "Real World"],train=True, transform=transform)
elif dataset_name == "OfficeHomeArt":
trainset = OfficeHome(root, domains=["Art"], train=True, transform=transform)
valset = OfficeHome(root, domains=["Art"], train=False, transform=transform)
testset = OfficeHome(root, domains=["Clipart", "Product", "Real World"],train=True, transform=transform)
else:
raise ValueError(f"{dataset_name} Dataset not supported")
return trainset, valset, testset
DATASET_CLASSES = {
"Waterbirds": ['landbird', 'waterbird'],
"Waterbirds95": ['landbird', 'waterbird'],
"ColoredMNISTBinary": ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9'],
"MNIST": ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9'],
"SVHN": ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9'],
"MNIST_SVHN": ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9'],
"DomainNet": DOMAINNET_CLASSES,
"DomainNetMini": MINI_DOMAINNET_CLASSES,
"DomainNetMiniOracle": MINI_DOMAINNET_CLASSES,
"CUB": CUB_CLASSES,
"OfficeHomeProduct": OFFICE_HOME_CLASSES,
"OfficeHomeClipart": OFFICE_HOME_CLASSES,
"OfficeHomeArt": OFFICE_HOME_CLASSES,
}
DATASET_DOMAINS = {
"Waterbirds": ['forest', 'water'],
"Waterbirds95": ['forest', 'water'],
"ColoredMNISTBinary": ['red', 'blue'],
"DomainNetMini": MINI_DOMAINS,
"DomainNetMiniOracle": MINI_DOMAINS,
"CUB": CUB_DOMAINS,
"SVHN": ["MNIST", "SVHN"],
"MNIST_SVHN": ["MNIST", "SVHN"],
"OfficeHomeProduct": OFFICE_HOME_DOMAINS,
"OfficeHomeClipart": OFFICE_HOME_DOMAINS,
"OfficeHomeArt": OFFICE_HOME_DOMAINS,
}
def get_domain(dataset_name):
return DATASET_DOMAINS[dataset_name]
def get_class(dataset_name):
return DATASET_CLASSES[dataset_name]