-
Notifications
You must be signed in to change notification settings - Fork 153
/
data_module.py
148 lines (122 loc) · 4.71 KB
/
data_module.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
# Copyright (c) 2021, Zhiqiang Wang. All Rights Reserved.
from pathlib import Path
import torch.utils.data
from torch.utils.data.dataset import Dataset
from pytorch_lightning import LightningDataModule
from typing import Callable, List, Any, Optional
from .transforms import collate_fn, default_train_transforms, default_val_transforms
from .voc import VOCDetection
from .coco import COCODetection
class DetectionDataModule(LightningDataModule):
"""
Wrapper of Datasets in LightningDataModule
"""
def __init__(
self,
train_dataset: Optional[Dataset] = None,
val_dataset: Optional[Dataset] = None,
test_dataset: Optional[Dataset] = None,
batch_size: int = 1,
num_workers: int = 0,
*args: Any,
**kwargs: Any,
) -> None:
super().__init__(*args, **kwargs)
self._train_dataset = train_dataset
self._val_dataset = val_dataset
self._test_dataset = test_dataset
self.batch_size = batch_size
self.num_workers = num_workers
def train_dataloader(self, batch_size: int = 16) -> None:
"""
VOCDetection and COCODetection
Args:
batch_size: size of batch
transforms: custom transforms
"""
# Creating data loaders
sampler = torch.utils.data.RandomSampler(self._train_dataset)
batch_sampler = torch.utils.data.BatchSampler(sampler, batch_size, drop_last=True)
loader = torch.utils.data.DataLoader(
self._train_dataset,
batch_sampler=batch_sampler,
collate_fn=collate_fn,
num_workers=self.num_workers,
)
return loader
def val_dataloader(self, batch_size: int = 16) -> None:
"""
VOCDetection and COCODetection
Args:
batch_size: size of batch
transforms: custom transforms
"""
# Creating data loaders
sampler = torch.utils.data.SequentialSampler(self._val_dataset)
loader = torch.utils.data.DataLoader(
self._val_dataset,
batch_size,
sampler=sampler,
drop_last=False,
collate_fn=collate_fn,
num_workers=self.num_workers,
)
return loader
class COCODetectionDataModule(DetectionDataModule):
def __init__(
self,
data_path: str,
year: str = "2017",
train_transform: Optional[Callable] = default_train_transforms,
val_transform: Optional[Callable] = default_val_transforms,
batch_size: int = 1,
num_workers: int = 0,
*args: Any,
**kwargs: Any,
) -> None:
train_dataset = self.build_datasets(
data_path, image_set='train', year=year, transforms=train_transform)
val_dataset = self.build_datasets(
data_path, image_set='val', year=year, transforms=val_transform)
super().__init__(train_dataset=train_dataset, val_dataset=val_dataset,
batch_size=batch_size, num_workers=num_workers, *args, **kwargs)
self.num_classes = 80
@staticmethod
def build_datasets(data_path, image_set, year, transforms):
ann_file = Path(data_path) / 'annotations' / f"instances_{image_set}{year}.json"
return COCODetection(data_path, ann_file, transforms())
class VOCDetectionDataModule(DetectionDataModule):
def __init__(
self,
data_path: str,
years: List[str] = ["2007", "2012"],
train_transform: Optional[Callable] = default_train_transforms,
val_transform: Optional[Callable] = default_val_transforms,
batch_size: int = 1,
num_workers: int = 0,
*args: Any,
**kwargs: Any,
) -> None:
train_dataset, num_classes = self.build_datasets(
data_path, image_set='train', years=years, transforms=train_transform)
val_dataset, _ = self.build_datasets(
data_path, image_set='val', years=years, transforms=val_transform)
super().__init__(train_dataset=train_dataset, val_dataset=val_dataset,
batch_size=batch_size, num_workers=num_workers, *args, **kwargs)
self.num_classes = num_classes
@staticmethod
def build_datasets(data_path, image_set, years, transforms):
datasets = []
for year in years:
dataset = VOCDetection(
data_path,
year=year,
image_set=image_set,
transforms=transforms(),
)
datasets.append(dataset)
num_classes = len(datasets[0].prepare.CLASSES)
if len(datasets) == 1:
return datasets[0], num_classes
else:
return torch.utils.data.ConcatDataset(datasets), num_classes