diffusion_from_scratch/data.py at main · jamie-mcg/diffusion_from_scratch · 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
import torch
import torchvision
import torchvision.transforms as T
from torch.utils.data import Subset
import numpy as np


class CIFAR10Dataset:
    def __init__(
        self,
        dataset_path,
        crop_box=(25, 50, 25 + 128, 50 + 128),
        image_size=128,
        valid_split=0.05,
    ):
        self.dataset_path = dataset_path
        self.crop_box = crop_box
        self.image_size = image_size
        self.valid_split = valid_split

        self.transform = T.Compose(
            [
                T.RandomHorizontalFlip(),
                T.ToTensor(),
                T.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True),
            ]
        )
        self.transform_test = T.Compose(
            [
                T.ToTensor(),
                T.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True),
            ]
        )

        self.train_set = torchvision.datasets.CIFAR10(
            self.dataset_path, train=True, transform=self.transform, download=True
        )
        self.valid_set = torchvision.datasets.CIFAR10(
            self.dataset_path,
            train=False,
            transform=self.transform_test,
            download=True,
        )

        self._create_train_valid_split()

    def _create_train_valid_split(self):
        num_train = len(self.train_set)
        indices = torch.randperm(num_train).tolist()
        valid_size = int(np.floor(self.valid_split * num_train))

        train_idx, valid_idx = indices[valid_size:], indices[:valid_size]

        self.train_set = Subset(self.train_set, train_idx)
        self.valid_set = Subset(self.valid_set, valid_idx)

    def get_train_set(self):
        return self.train_set

    def get_valid_set(self):
        return self.valid_set


class CelebADataset:
    def __init__(
        self,
        dataset_path,
        crop_box=(25, 50, 25 + 128, 50 + 128),
        image_size=128,
        valid_split=0.05,
    ):
        self.dataset_path = dataset_path
        self.crop_box = crop_box
        self.image_size = image_size
        self.valid_split = valid_split

        self.transform = T.Compose(
            [
                T.Lambda(lambda img: img.crop(self.crop_box)),
                T.Resize(self.image_size),
                T.RandomHorizontalFlip(),
                T.ToTensor(),
                T.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True),
            ]
        )

        self.transform_test = T.Compose(
            [
                T.Lambda(lambda img: img.crop(self.crop_box)),
                T.Resize(self.image_size),
                T.ToTensor(),
                T.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True),
            ]
        )

        self.train_set = torchvision.datasets.CelebA(
            self.dataset_path, train=True, transform=self.transform, download=True
        )
        self.valid_set = torchvision.datasets.CelebA(
            self.dataset_path,
            train=False,
            transform=self.transform_test,
            download=True,
        )

        self._create_train_valid_split()

    def _create_train_valid_split(self):
        num_train = len(self.train_set)
        indices = torch.randperm(num_train).tolist()
        valid_size = int(np.floor(self.valid_split * num_train))

        train_idx, valid_idx = indices[valid_size:], indices[:valid_size]

        self.train_set = Subset(self.train_set, train_idx)
        self.valid_set = Subset(self.valid_set, valid_idx)

    def get_train_set(self):
        return self.train_set

    def get_valid_set(self):
        return self.valid_set


if __name__ == "__main__":
    import os
    import torch

    DATA_PATH = os.environ.get("DATA_PATH")

    dataset = CIFAR10Dataset(dataset_path=DATA_PATH)