-
Notifications
You must be signed in to change notification settings - Fork 16
/
dataloader.py
144 lines (121 loc) · 4.79 KB
/
dataloader.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
import numpy as np
import scipy.io as sio
from skimage import io
from torch.utils.data import Dataset, DataLoader
import torch
from torchvision import transforms
import augmentation
from PIL import Image
class ImageData:
def __init__(self):
self.cropped_image_path = ''
self.cropped_posmap_path = ''
self.init_image_path = ''
self.init_posmap_path = ''
self.texture_path = ''
self.texture_image_path = ''
self.bbox_info_path = ''
self.offset_posmap_path = ''
self.attention_mask_path = ''
self.image = None
self.posmap = None
self.offset_posmap = None
self.bbox_info = None
self.S = None
self.T = None
self.R = None
self.attention_mask = None
def readPath(self, image_dir):
image_name = image_dir.split('/')[-1]
self.cropped_image_path = image_dir + '/' + image_name + '_cropped.jpg'
self.cropped_posmap_path = image_dir + '/' + image_name + '_cropped_uv_posmap.npy'
self.init_image_path = image_dir + '/' + image_name + '_init.jpg'
self.init_posmap_path = image_dir + '/' + image_name + '_uv_posmap.npy'
# change the format to npy
self.texture_path = image_dir + '/' + image_name + '_uv_texture_map.npy'
self.texture_image_path = image_dir + '/' + image_name + '_uv_texture_map.jpg'
self.bbox_info_path = image_dir + '/' + image_name + '_bbox_info.mat'
def readFile(self, mode='posmap'):
if mode == 'posmap':
self.image = io.imread(self.cropped_image_path).astype(np.uint8)
self.posmap = np.load(self.cropped_posmap_path).astype(np.float16)
else:
pass
def getImage(self):
if self.image is None:
return io.imread(self.cropped_image_path)
else:
return self.image
def getPosmap(self):
if self.posmap is None:
return np.load(self.cropped_posmap_path)
else:
return self.posmap
def getOffsetPosmap(self):
if self.offset_posmap is None:
return np.load(self.offset_posmap_path)
else:
return self.offset_posmap
def getBboxInfo(self):
if self.bbox_info is None:
return sio.loadmat(self.bbox_info_path)
else:
return self.bbox_info
def getAttentionMask(self):
if self.attention_mask is None:
return np.load(self.attention_mask_path)
else:
return self.attention_mask
def toTensor(image):
image = image.transpose((2, 0, 1))
image = torch.from_numpy(image)
return image
class DataGenerator(Dataset):
def __init__(self, all_image_data, mode='posmap', is_aug=False, is_pre_read=True):
super(DataGenerator, self).__init__()
self.all_image_data = all_image_data
self.image_height = 256
self.image_width = 256
self.image_channel = 3
# mode=posmap or offset
self.mode = mode
self.is_aug = is_aug
self.toTensor = transforms.ToTensor()
self.is_pre_read = is_pre_read
if is_pre_read:
i = 0
print('preloading')
num_max_PR = 80000
for data in self.all_image_data:
data.readFile(mode=self.mode)
print(i, end='\r')
i += 1
if i > num_max_PR:
break
def __getitem__(self, index):
if self.mode == 'posmap':
image = (self.all_image_data[index].getImage() / 255.0).astype(np.float32)
pos = self.all_image_data[index].getPosmap().astype(np.float32)
# data augmentation
if self.is_aug:
image, pos = augmentation.prnAugment_torch(image, pos)
for i in range(3):
image[:, :, i] = (image[:, :, i] - image[:, :, i].mean()) / np.sqrt(image[:, :, i].var() + 0.001)
image = self.toTensor(image)
else:
for i in range(3):
image[:, :, i] = (image[:, :, i] - image[:, :, i].mean()) / np.sqrt(image[:, :, i].var() + 0.001)
image = self.toTensor(image)
pos = pos / 280.
pos = self.toTensor(pos)
return image, pos
else:
import os
os.error('please use "posmap" mode')
return None
def __len__(self):
return len(self.all_image_data)
def getDataLoader(all_image_data, mode='posmap', batch_size=16, is_shuffle=False, is_aug=False, is_pre_read=True, num_worker=8):
dataset = DataGenerator(all_image_data=all_image_data, mode=mode, is_aug=is_aug, is_pre_read=is_pre_read)
train_loader = DataLoader(dataset=dataset, batch_size=batch_size, shuffle=is_shuffle, num_workers=num_worker, pin_memory=True)
return train_loader