-
Notifications
You must be signed in to change notification settings - Fork 2
/
dataset.py
270 lines (223 loc) · 9.59 KB
/
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
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
import torch
import cv2
import os.path as osp
from torch.utils import data
import glob
from PIL import Image
import numpy as np
import random
import re
import os
import torchvision.transforms as transforms
import torchvision.transforms.functional as F
from skimage.color import rgb2yuv
def my_collate(batch):
imgs,targets,cvimgs = zip(*batch)
return torch.cat(imgs),torch.cat(targets),cvimgs
class ColorJitter(object):
def __init__(self,b=0.3,c=0.3,s=0.3,h=3.1415/6):
super(ColorJitter,self).__init__()
self.b = b
self.c = c
self.s = s
self.h = h
def __call__(self, img):
b_val = random.uniform(-self.b,self.b)
c_val = random.uniform(1-self.c,1+self.c)
s_val = random.uniform(1-self.s,1+self.s)
h_val = random.uniform(-self.h,self.h)
mtx = torch.FloatTensor([[s_val*np.cos(h_val),-np.sin(h_val)],[np.sin(h_val),s_val*np.cos(h_val)]])
img[0] = (img[0]+b_val)*c_val
if self.s > 0 and self.h > 0:
img[1:] = torch.einsum('nm,mbc->nbc',mtx,img[1:])
return img
def tryint(s):
try:
return int(s)
except:
return s
def alphanum_key(s):
""" Turn a string into a list of string and number chunks.
"z23a" -> ["z", 23, "a"]
"""
return [ tryint(c) for c in re.split('([0-9]+)', s) ]
def default_loader(path):
return Image.open(path)
def get_immediate_subdirectories(a_dir):
return [name for name in os.listdir(a_dir)
if os.path.isdir(os.path.join(a_dir, name))]
class ToYUV(object):
def __call__(self, img):
return rgb2yuv(img)
#return img.convert('YCbCr')
class SSYUVDataset(data.Dataset):
def __init__(self, data_dir, img_size=(120,160), train=True, finetune = False,camera = "both"):
self.img_shape = img_size
self.train = train
self.finetune = finetune
self.img_size = img_size
self.jitter = ColorJitter(0.3,0.3,0.3,3.1415/6)
self.resize = transforms.Resize(img_size)
self.labResize = transforms.Resize(img_size,Image.NEAREST)
self.mean = [0.34190056, 0.4833289, 0.48565758] if finetune else [0.36269532, 0.41144562, 0.282713]
self.std = [0.47421749, 0.13846053, 0.1714848] if finetune else [0.31111388, 0.21010718, 0.34060917]
self.normalize = transforms.Normalize(mean=self.mean,std=self.std)
self.images =[]
self.labels =[]
if finetune:
data_dir = osp.join(data_dir,"FinetuneHorizon")
data_dir = osp.join(data_dir,"train" if train else "val")
self.img_dir = osp.join(data_dir, "images")
self.lab_dir = osp.join(data_dir, "labels")
imgFiles = sorted(glob.glob1(self.img_dir, "*.png"), key=alphanum_key)
txtFiles = sorted(glob.glob1(self.img_dir, "*.txt"), key=alphanum_key)
labFiles = sorted(glob.glob1(self.lab_dir, "*.png"), key=alphanum_key)
if len(txtFiles) == len(imgFiles):
for img, lab, txt in zip(imgFiles, labFiles, txtFiles):
char = open(osp.join(self.img_dir, txt)).read()
condition = (camera == "both") or ((camera == "top") and (char == "u")) or (
(camera == "bottom") and (char == "b"))
if condition:
self.images.append(img)
self.labels.append(lab)
else:
for img, lab in zip(imgFiles, labFiles):
self.images.append(img)
self.labels.append(lab)
def __len__(self):
return len(self.images)
def __getitem__(self, index):
#---------
# Image
#---------
img_file = osp.join(self.img_dir, self.images[index])
lab_file = osp.join(self.lab_dir, self.labels[index])
img = Image.open(img_file).convert('RGB')
label = Image.open(lab_file).convert('I')
if self.img_size[0] != img.size[1] and self.img_size[1] != img.size[0]:
img = self.resize(img)
if self.img_size[0] != label.size[1] and self.img_size[1] != label.size[0]:
label = self.labResize(label)
img = transforms.functional.to_tensor(img).float()
label = transforms.functional.to_tensor(label)
img = self.normalize(img)
if self.train:
p = torch.rand(1).item()
if p > 0.5:
img = img.flip(2)
label = label.flip(2)
img = self.jitter(img)
return img, label.squeeze()
class SSDataSet(data.Dataset):
def __init__(self, root, split="train", camera = "both", img_transform=None, label_transform=None):
self.root = root
self.split = split
self.images = []
self.labels = []
self.labels = []
self.img_transform = img_transform
self.label_transform = label_transform
data_dir = osp.join(root, split)
self.img_dir = osp.join(data_dir,"images")
self.lab_dir = osp.join(data_dir,"labels")
imgFiles = sorted(glob.glob1(self.img_dir, "*.png"),key=alphanum_key)
txtFiles = sorted(glob.glob1(self.img_dir, "*.txt"),key=alphanum_key)
labFiles = sorted(glob.glob1(self.lab_dir, "*.png"),key=alphanum_key)
if len(txtFiles) == len(imgFiles):
for img,lab,txt in zip(imgFiles,labFiles,txtFiles):
char = open(osp.join( self.img_dir, txt )).read()
condition = (camera== "both") or ((camera == "top") and (char == "u")) or ((camera == "bottom") and (char == "b"))
if condition:
self.images.append(img)
self.labels.append(lab)
else:
for img,lab in zip(imgFiles,labFiles):
self.images.append(img)
self.labels.append(lab)
def __len__(self):
return len(self.images)
def __getitem__(self, index):
img_file = osp.join( self.img_dir, self.images[index])
lab_file = osp.join( self.lab_dir, self.labels[index])
img = Image.open(img_file).convert('RGB')
label = Image.open(lab_file).convert("I")
seed = np.random.randint(2147483647) # make a seed with numpy generator
random.seed(seed) # apply this seed to img tranfsorms
if self.img_transform is not None:
imgs = self.img_transform(img)
else:
imgs = img
random.seed(seed) # apply this seed to target tranfsorms
if self.label_transform is not None:
labels = self.label_transform(label)
else:
labels = label
return imgs, labels
class LPDataSet(data.Dataset):
def __init__(self, root, train=True, img_size=(120,160), finetune=True, len_seq = 2):
self.finetune = finetune
self.img_size = img_size
self.len_seq = len_seq
self.root = osp.join(root,"LabelProp")
self.split = "train" if train else "val"
self.resize = transforms.Resize(img_size)
self.labResize = transforms.Resize(img_size,Image.NEAREST)
self.mean = [0.34190056, 0.4833289, 0.48565758] if finetune else [0.36269532, 0.41144562, 0.282713]
self.std = [0.47421749, 0.13846053, 0.1714848] if finetune else [0.31111388, 0.21010718, 0.34060917]
self.normalize = transforms.Normalize(mean=self.mean,std=self.std)
self.images = []
self.labels = []
self.predictions = []
data_dir = osp.join(self.root,"Real" if finetune else "Synthetic")
data_dir = osp.join(data_dir, self.split)
for dir in get_immediate_subdirectories(data_dir):
currDir = osp.join(data_dir,dir)
img_dir = osp.join(currDir,"images")
lab_dir = osp.join(currDir,"labels")
images = []
labels = []
for file in sorted(glob.glob1(img_dir, "*.png"), key=alphanum_key):
images.append(osp.join(img_dir, file))
for file in sorted(glob.glob1(lab_dir, "*.png"), key=alphanum_key):
labels.append(osp.join(lab_dir,file))
self.images.append(images)
self.labels.append(labels)
def __len__(self):
length = 0
for imgs in self.images:
length += len(imgs) - self.len_seq + 1
return length
def __getitem__(self, index):
dirindex = 0
itemindex = index
#print index
for imgs in self.images:
#print(dirindex, itemindex, len(imgs))
if itemindex >= len(imgs) - self.len_seq + 1:
dirindex += 1
itemindex -= (len(imgs) - self.len_seq + 1)
else:
break
#print(dirindex, itemindex)
labels = []
imgs = []
cvimgs = []
for i in range(self.len_seq):
img_file = self.images[dirindex][itemindex+i]
lab_file = self.labels[dirindex][itemindex+i]
img = Image.open(img_file).convert('RGB')
label = Image.open(lab_file).convert("I")
if self.img_size[0] != img.size[1] and self.img_size[1] != img.size[0]:
img = self.resize(img)
if self.img_size[0] != label.size[1] and self.img_size[1] != label.size[0]:
label = self.labResize(label)
img_ten = cv2.cvtColor(np.array(img),cv2.COLOR_RGB2YUV)
img_ten = transforms.functional.to_tensor(img_ten).float()
label = transforms.functional.to_tensor(label)
img_ten = self.normalize(img_ten).unsqueeze(0)
imgs.append(img_ten)
cvimgs.append(cv2.cvtColor(np.array(img),cv2.COLOR_RGB2GRAY))
labels.append(label)
imgs = torch.cat(imgs)
labels = torch.cat(labels)
return imgs, labels, cvimgs