-
Notifications
You must be signed in to change notification settings - Fork 0
/
data.py
143 lines (114 loc) · 4.07 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
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
import numpy as np
import pandas as pd
from typing import Dict, List, Tuple
import json
import pathlib
from PIL import Image
import os
from pytorch_detection.utils import collate_fn
from pytorch_detection import utils
from pytorch_detection import transforms as T
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
from torchvision import datasets, models, transforms
import pytorch_detection.transforms as T
from torch.utils.data.dataset import Dataset
from torch.utils.data import DataLoader
device = torch.device("cuda")
class DataHandler:
def __init__(self, run_config):
self._training_dataset = None
self._validation_dataset = None
self._run_config = run_config
self.root_path = "./dataset"
self.load_datasets()
def load_datasets(self):
masked_faces_paths = list(pathlib.Path(self.root_path + "/masked_faces").glob('*'))
normal_faces_paths = list(pathlib.Path(self.root_path + "/normal_faces").glob('*'))
ratio = 90
length = len(masked_faces_paths)
training_size = int(0.8 * length)
training_paths = masked_faces_paths[:training_size] + normal_faces_paths[:training_size]
validation_paths = masked_faces_paths[training_size:] + normal_faces_paths[training_size:]
self._training_dataset = CustomDataset(training_paths, root_path="./dataset", run_type="train")
self._validation_dataset = CustomDataset(validation_paths, root_path="./dataset", run_type="valid")
def get_data_loaders(self) -> Tuple[DataLoader]:
return (
DataLoader(
self._training_dataset,
batch_size=self._run_config.batch_size,
shuffle=True,
num_workers=self._run_config.workers,
pin_memory=False,
collate_fn=collate_fn
),
DataLoader(
self._validation_dataset,
batch_size=self._run_config.batch_size,
shuffle=True,
num_workers=self._run_config.workers,
pin_memory=False,
collate_fn=collate_fn
)
)
def get_datasets(self) -> Tuple[Dataset]:
return self._training_dataset, self._validation_dataset
def get_datasets_sizes(self) -> Tuple[int]:
return len(self._training_dataset), len(self._validation_dataset)
class CustomDataset(Dataset):
def __init__(self, image_paths, root_path = "./dataset", run_type = "train"):
self.root_path = root_path
self.run_type = run_type
with open(self.root_path + "/targets.json") as json_file:
self.targets = json.load(json_file)
self.images_paths = image_paths
self.transformers = {
'train_transforms' : T.Compose([
# transforms.RandomHorizontalFlip(0.5),
T.ToTensor()
]),
'valid_transforms' : T.Compose([
T.ToTensor()
])
}
def __getitem__(self, idx):
image_path = self.images_paths[idx]
image_name = str(image_path).split(os.sep)[2]
image_class = str(image_path).split(os.sep)[1]
class_to_label = {
"masked_faces": 1,
"normal_faces": 2
}
label = class_to_label[image_class]
image = Image.open(image_path).convert("RGB")
target = {}
if len(self.targets[image_name]["bbox"]) == 0:
self.targets[image_name]["bbox"] = [[]]
target["boxes"] = torch.as_tensor(self.targets[image_name]["bbox"], dtype=torch.float32).to(device)
target["labels"] = torch.as_tensor([label] * len(self.targets[image_name]["bbox"]), dtype=torch.int64).to(device)
target["iscrowd"] = torch.tensor([0] * len(self.targets[image_name]["bbox"]), dtype=torch.int64).to(device)
target["image_id"] = torch.tensor([idx]).to(device)
area = (
target["boxes"][:, 3] - \
target["boxes"][:, 1]) * \
(target["boxes"][:, 2] - \
target["boxes"][:, 0]
)
target["area"] = torch.tensor(area)
if self.transformers is not None:
image, target = self.transformers[self.run_type + "_transforms"](image, target)
return image, target
def __len__(self):
return len(self.images_paths)
def get_aug(aug, min_area=0., min_visibility=0.):
return Compose(
aug,
bbox_params=BboxParams(
format='pascal_voc',
min_area=min_area,
min_visibility=min_visibility,
label_fields=['category_id']
)
)