-
Notifications
You must be signed in to change notification settings - Fork 0
/
datasets.py
134 lines (110 loc) · 4.12 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
from torchvision import datasets, transforms
from torch.utils.data import DataLoader
import torch
import os
import csv
import PIL
from collections import namedtuple
from utils import kaggle_setup
from zipfile import ZipFile
FROM_KAGGLE = False
CSV = namedtuple("CSV", ["header", "index", "data"])
def crop_celeba(img):
return transforms.functional.crop(img, top=40, left=15, height=148, width=148)
class CelebA(datasets.VisionDataset):
"""
Large-scale CelebFaces Attributes (CelebA) Dataset <http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html>
"""
base_folder = "celeba-kaggle"
def __init__(
self,
root,
split,
download=False,
transform=transforms.ToTensor()
):
super().__init__(root, transform=transform)
self.split = split
split_map = {
"train": 0,
"valid": 1,
"test": 2,
"all": None,
}
if download:
kaggle_setup()
download_folder = os.path.join(self.root, self.base_folder)
kaggle_ref = "jessicali9530/celeba-dataset"
os.system(f"kaggle datasets download -p {download_folder} {kaggle_ref}")
print("Decompressing the downloaded file...")
file = os.path.join(download_folder, kaggle_ref.split("/")[-1] + ".zip")
# os.system(f"unzip -q -d {download_folder} {file}") # too slow
with ZipFile(file, "r") as zf:
zf.extractall(path=download_folder)
split_ = split_map[split.lower()]
splits = self._load_csv("list_eval_partition.csv", header=0)
mask = slice(None) if split_ is None else (splits.data == split_).squeeze()
if mask == slice(None): # if split == "all"
self.filename = splits.index
else:
self.filename = [splits.index[i] for i in torch.squeeze(torch.nonzero(mask))]
self.download = download
def _load_csv(
self,
filename,
header=None,
):
with open(os.path.join(self.root, self.base_folder, filename)) as csv_file:
data = list(csv.reader(csv_file, delimiter=",", skipinitialspace=True))
if header is not None:
headers = data[header]
data = data[header + 1:]
else:
headers = []
indices = [row[0] for row in data]
data = [row[1:] for row in data]
data_int = [list(map(int, i)) for i in data]
return CSV(headers, indices, torch.tensor(data_int))
def __getitem__(self, index):
X = PIL.Image.open(os.path.join(
self.root, self.base_folder, "img_align_celeba", "img_align_celeba", self.filename[index]))
if self.transform is not None:
X = self.transform(X)
return X, 0
def __len__(self):
return len(self.filename)
def extra_repr(self):
lines = ["Split: {split}", ]
return "\n".join(lines).format(**self.__dict__)
transform_mnist = transforms.Compose([
transforms.Resize((32, 32)),
transforms.ToTensor()
])
transform_cifar10 = transforms.Compose([
transforms.Resize((32, 32)),
transforms.RandomHorizontalFlip(p=0.5),
transforms.ToTensor()
])
transform_celeba = transforms.Compose([
crop_celeba,
transforms.Resize((64, 64)),
transforms.RandomHorizontalFlip(p=0.5),
transforms.ToTensor()
])
DATASETS = {
"mnist": datasets.MNIST,
"cifar10": datasets.CIFAR10,
"celeba": CelebA if FROM_KAGGLE else datasets.CelebA
}
def get_data(root, dataset, download, batch_size):
dataset = DATASETS[dataset]
if dataset == "mnist":
train_data = dataset(root=root, download=download, train=True, transform=transform_mnist)
elif dataset == "cifar10":
train_data = dataset(root=root, download=download, train=True, transform=transform_cifar10)
elif dataset == "celeba":
train_data = dataset(root=root, download=download, split="all", transform=transform_celeba)
else:
raise NotImplementedError
train_loader = DataLoader(train_data, batch_size, num_workers=os.cpu_count(), shuffle=True, pin_memory=True)
return train_loader