/
data.py
69 lines (56 loc) · 2.7 KB
/
data.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
from omegaconf import DictConfig
from hydra.utils import to_absolute_path
from pytorch_lightning import LightningDataModule
import os
import torch
import torchvision.transforms as T
from torchvision.datasets import ImageFolder
import torchvision.io as io
from functools import partial
class DataModule(LightningDataModule):
def __init__(self, cfg: DictConfig):
super().__init__()
# Hydra changes working directory, so change cfg.data.root to
# reflect this change
cfg.data.root = to_absolute_path(cfg.data.root)
cfg.data.root = os.path.abspath(cfg.data.root)
self.cfg = cfg.data
self.dataset_path = os.path.join(self.cfg.root, self.cfg.name)
valid_names = ("imagenette2", "imagewoof2")
if self.cfg.name not in valid_names:
raise ValueError(f"Incorrect \"data.name: {self.cfg.name}. The "
f"valid options are {valid_names}")
if cfg.apply_resizer_model:
image_size = self.cfg.resizer_image_size
else:
image_size = self.cfg.image_size
self.image_read_func = partial(io.read_image,
mode=io.image.ImageReadMode.RGB)
self.train_transform = T.Compose([
T.Resize((image_size, image_size)),
T.RandomHorizontalFlip(),
T.ConvertImageDtype(torch.float32),
T.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
])
self.test_transform = T.Compose([
T.Resize((image_size, image_size)),
T.ConvertImageDtype(torch.float32),
T.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
])
def setup(self, stage=None):
self.train_data = ImageFolder(os.path.join(self.dataset_path, 'train'),
transform=self.train_transform,
loader=self.image_read_func)
self.val_data = ImageFolder(os.path.join(self.dataset_path, 'val'),
transform=self.test_transform,
loader=self.image_read_func)
self.val_length = len(self.val_data)
def train_dataloader(self):
return torch.utils.data.DataLoader(dataset=self.train_data,
batch_size=self.cfg.batch_size,
shuffle=True,
num_workers=self.cfg.num_workers)
def val_dataloader(self):
return torch.utils.data.DataLoader(dataset=self.val_data,
batch_size=self.cfg.batch_size,
num_workers=self.cfg.num_workers)