-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathgeneric_dataset.py
158 lines (144 loc) · 6.06 KB
/
generic_dataset.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
import logging
import sys
import numpy as np
import torch
from .basedataset import BaseDataset
sys.path.append("../")
from PIL import Image
from torch.utils.data import Dataset
logging.getLogger("PIL").setLevel(logging.WARNING)
class GenericImageExpertDataset(Dataset):
def __init__(self, images, targets, expert_preds, transforms_fn, to_open=False):
"""
Args:
images (list): List of images
targets (list): List of labels
expert_preds (list): List of expert predictions
transforms_fn (function): Function to apply to images
to_open (bool): Whether to open images or not (RGB reader)
"""
self.images = images
self.targets = np.array(targets)
self.expert_preds = np.array(expert_preds)
self.transforms_fn = transforms_fn
self.to_open = to_open
def __getitem__(self, index):
"""Take the index of item and returns the image, label, expert prediction and index in original dataset"""
label = self.targets[index]
if self.transforms_fn is not None and self.to_open:
image_paths = self.images[index]
image = Image.open(image_paths).convert("RGB")
image = self.transforms_fn(image)
elif self.transforms_fn is not None:
image = self.transforms_fn(self.images[index])
else:
image = self.images[index]
expert_pred = self.expert_preds[index]
return torch.FloatTensor(image), label, expert_pred
def __len__(self):
return len(self.targets)
class GenericDatasetDeferral(BaseDataset):
def __init__(
self,
data_train,
data_test=None,
test_split=0.2,
val_split=0.1,
batch_size=100,
transforms=None,
):
"""
data_train: training data expectd as dict with keys 'data_x', 'data_y', 'hum_preds'
data_test: test data expectd as dict with keys 'data_x', 'data_y', 'hum_preds'
test_split: fraction of training data to use for test
val_split: fraction of training data to use for validation
batch_size: batch size for dataloaders
transforms: transforms to apply to images
"""
self.data_train = data_train
self.data_test = data_test
self.test_split = test_split
self.val_split = val_split
self.batch_size = batch_size
self.train_split = 1 - test_split - val_split
self.transforms = transforms
self.generate_data()
def generate_data(self):
train_x = self.data_train["data_x"]
train_y = self.data_train["data_y"]
train_hum_preds = self.data_train["hum_preds"]
if self.data_test is not None:
test_x = self.data_test["data_x"]
test_y = self.data_test["data_y"]
test_h = self.data_test["hum_preds"]
train_size = int((1 - self.val_split) * self.total_samples)
val_size = int(self.val_split * self.total_samples)
train_x, val_x = torch.utils.data.random_split(
train_x,
[train_size, val_size],
generator=torch.Generator().manual_seed(42),
)
train_y, val_y = torch.utils.data.random_split(
train_y,
[train_size, val_size],
generator=torch.Generator().manual_seed(42),
)
train_h, val_h = torch.utils.data.random_split(
train_hum_preds,
[train_size, val_size],
generator=torch.Generator().manual_seed(42),
)
self.data_train = torch.utils.data.TensorDataset(
train_x.dataset.data[train_x.indices],
train_y.dataset.data[train_y.indices],
train_h.dataset.data[train_h.indices],
)
self.data_val = torch.utils.data.TensorDataset(
val_x.dataset.data[val_x.indices],
val_y.dataset.data[val_y.indices],
val_h.dataset.data[val_h.indices],
)
self.data_test = torch.utils.data.TensorDataset(test_x, test_y, test_h)
else:
train_size = int(self.train_split * self.total_samples)
val_size = int(self.val_split * self.total_samples)
test_size = self.total_samples - train_size - val_size
train_x, val_x, test_x = torch.utils.data.random_split(
train_x,
[train_size, val_size, test_size],
generator=torch.Generator().manual_seed(42),
)
train_y, val_y, test_y = torch.utils.data.random_split(
train_y,
[train_size, val_size, test_size],
generator=torch.Generator().manual_seed(42),
)
train_h, val_h, test_h = torch.utils.data.random_split(
train_hum_preds,
[train_size, val_size, test_size],
generator=torch.Generator().manual_seed(42),
)
self.data_train = torch.utils.data.TensorDataset(
train_x.dataset.data[train_x.indices],
train_y.dataset.data[train_y.indices],
train_h.dataset.data[train_h.indices],
)
self.data_val = torch.utils.data.TensorDataset(
val_x.dataset.data[val_x.indices],
val_y.dataset.data[val_y.indices],
val_h.dataset.data[val_h.indices],
)
self.data_test = torch.utils.data.TensorDataset(
test_x.dataset.data[test_x.indices],
test_y.dataset.data[test_y.indices],
test_h.dataset.data[test_h.indices],
)
self.data_train_loader = torch.utils.data.DataLoader(
self.data_train, batch_size=self.batch_size, shuffle=True
)
self.data_val_loader = torch.utils.data.DataLoader(
self.data_val, batch_size=self.batch_size, shuffle=True
)
self.data_test_loader = torch.utils.data.DataLoader(
self.data_test, batch_size=self.batch_size, shuffle=True
)