-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathutils.py
236 lines (205 loc) · 9.05 KB
/
utils.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
# Copyright Niantic 2019. Patent Pending. All rights reserved.
#
# This software is licensed under the terms of the Monodepth2 licence
# which allows for non-commercial use only, the full terms of which are made
# available in the LICENSE file.
from __future__ import absolute_import, division, print_function
import os, shutil
import hashlib
import zipfile
import numpy as np
from six.moves import urllib
def save_code(srcfile, log_path):
#save depth_decoder.py to log, easy to evaluation
#srcfile = "./networks/depth_decoder.py"
if not os.path.isfile(srcfile):
print ("%s not exist!"%(srcfile))
else:
fpath,fname=os.path.split(srcfile)
if not os.path.exists(log_path):
os.makedirs(log_path)
shutil.copy(srcfile, os.path.join(log_path, fname))
print ("copy %s -> %s"%(srcfile, os.path.join(log_path, fname)))
def readlines(filename):
"""Read all the lines in a text file and return as a list
"""
with open(filename, 'r') as f:
lines = f.read().splitlines()
return lines
def normalize_image(x):
"""Rescale image pixels to span range [0, 1]
"""
ma = float(x.max().cpu().data)
mi = float(x.min().cpu().data)
d = ma - mi if ma != mi else 1e5
return (x - mi) / d
def sec_to_hm(t):
"""Convert time in seconds to time in hours, minutes and seconds
e.g. 10239 -> (2, 50, 39)
"""
t = int(t)
s = t % 60
t //= 60
m = t % 60
t //= 60
return t, m, s
def sec_to_hm_str(t):
"""Convert time in seconds to a nice string
e.g. 10239 -> '02h50m39s'
"""
h, m, s = sec_to_hm(t)
return "{:02d}h{:02d}m{:02d}s".format(h, m, s)
def download_model_if_doesnt_exist(model_name):
"""If pretrained kitti model doesn't exist, download and unzip it
"""
# values are tuples of (<google cloud URL>, <md5 checksum>)
download_paths = {
"mono_640x192":
("https://storage.googleapis.com/niantic-lon-static/research/monodepth2/mono_640x192.zip",
"a964b8356e08a02d009609d9e3928f7c"),
"stereo_640x192":
("https://storage.googleapis.com/niantic-lon-static/research/monodepth2/stereo_640x192.zip",
"3dfb76bcff0786e4ec07ac00f658dd07"),
"mono+stereo_640x192":
("https://storage.googleapis.com/niantic-lon-static/research/monodepth2/mono%2Bstereo_640x192.zip",
"c024d69012485ed05d7eaa9617a96b81"),
"mono_no_pt_640x192":
("https://storage.googleapis.com/niantic-lon-static/research/monodepth2/mono_no_pt_640x192.zip",
"9c2f071e35027c895a4728358ffc913a"),
"stereo_no_pt_640x192":
("https://storage.googleapis.com/niantic-lon-static/research/monodepth2/stereo_no_pt_640x192.zip",
"41ec2de112905f85541ac33a854742d1"),
"mono+stereo_no_pt_640x192":
("https://storage.googleapis.com/niantic-lon-static/research/monodepth2/mono%2Bstereo_no_pt_640x192.zip",
"46c3b824f541d143a45c37df65fbab0a"),
"mono_1024x320":
("https://storage.googleapis.com/niantic-lon-static/research/monodepth2/mono_1024x320.zip",
"0ab0766efdfeea89a0d9ea8ba90e1e63"),
"stereo_1024x320":
("https://storage.googleapis.com/niantic-lon-static/research/monodepth2/stereo_1024x320.zip",
"afc2f2126d70cf3fdf26b550898b501a"),
"mono+stereo_1024x320":
("https://storage.googleapis.com/niantic-lon-static/research/monodepth2/mono%2Bstereo_1024x320.zip",
"cdc5fc9b23513c07d5b19235d9ef08f7"),
}
if not os.path.exists("models"):
os.makedirs("models")
model_path = os.path.join("models", model_name)
def check_file_matches_md5(checksum, fpath):
if not os.path.exists(fpath):
return False
with open(fpath, 'rb') as f:
current_md5checksum = hashlib.md5(f.read()).hexdigest()
return current_md5checksum == checksum
# see if we have the model already downloaded...
if not os.path.exists(os.path.join(model_path, "encoder.pth")):
model_url, required_md5checksum = download_paths[model_name]
if not check_file_matches_md5(required_md5checksum, model_path + ".zip"):
print("-> Downloading pretrained model to {}".format(model_path + ".zip"))
urllib.request.urlretrieve(model_url, model_path + ".zip")
if not check_file_matches_md5(required_md5checksum, model_path + ".zip"):
print(" Failed to download a file which matches the checksum - quitting")
quit()
print(" Unzipping model...")
with zipfile.ZipFile(model_path + ".zip", 'r') as f:
f.extractall(model_path)
print(" Model unzipped to {}".format(model_path))
import collections
from torch._six import string_classes
import re
import torch
np_str_obj_array_pattern = re.compile(r'[SaUO]')
default_collate_err_msg_format = (
"default_collate: batch must contain tensors, numpy arrays, numbers, "
"dicts or lists; found {}")
def rmnone_collate(batch):
batch_new = []
for v in batch:
if v is not None:
batch_new.append(v)
batch = batch_new
if len(batch) == 0:
return None
else:
return default_collate(batch)
def default_collate(batch):
r"""Puts each data field into a tensor with outer dimension batch size"""
elem = batch[0]
elem_type = type(elem)
if isinstance(elem, torch.Tensor):
out = None
if torch.utils.data.get_worker_info() is not None:
# If we're in a background process, concatenate directly into a
# shared memory tensor to avoid an extra copy
numel = sum(x.numel() for x in batch)
storage = elem.storage()._new_shared(numel)
out = elem.new(storage)
return torch.stack(batch, 0, out=out)
elif elem_type.__module__ == 'numpy' and elem_type.__name__ != 'str_' \
and elem_type.__name__ != 'string_':
if elem_type.__name__ == 'ndarray' or elem_type.__name__ == 'memmap':
# array of string classes and object
if np_str_obj_array_pattern.search(elem.dtype.str) is not None:
raise TypeError(default_collate_err_msg_format.format(elem.dtype))
return default_collate([torch.as_tensor(b) for b in batch])
elif elem.shape == (): # scalars
return torch.as_tensor(batch)
elif isinstance(elem, float):
return torch.tensor(batch, dtype=torch.float64)
elif isinstance(elem, int):
return torch.tensor(batch)
elif isinstance(elem, string_classes):
return batch
elif isinstance(elem, collections.abc.Mapping):
return {key: default_collate([d[key] for d in batch]) for key in elem}
elif isinstance(elem, tuple) and hasattr(elem, '_fields'): # namedtuple
return elem_type(*(default_collate(samples) for samples in zip(*batch)))
elif isinstance(elem, collections.abc.Sequence):
# check to make sure that the elements in batch have consistent size
it = iter(batch)
elem_size = len(next(it))
if not all(len(elem) == elem_size for elem in it):
raise RuntimeError('each element in list of batch should be of equal size')
transposed = zip(*batch)
return [default_collate(samples) for samples in transposed]
raise TypeError(default_collate_err_msg_format.format(elem_type))
def preprocess_image(image, delta=0.01, x_length=40, y_length=35):
# image should be float between [0, 1] with size HWC
new_image = image.copy()
H, W, C = image.shape
flaging = False
for y in range(H):
for x in range(W-1):
if np.abs((image[y, x] - image[y, x+1])).sum() < delta:
if not flaging:
flaging = True
start_x = x
else:
if flaging:
flaging = False
end_x = x
if (end_x - start_x) > x_length:
encoding = np.linspace(0, 1, end_x - start_x + 1)
# color = image[y, start_x:end_x+1, :].mean()
new_image[y, start_x:end_x+1, 0] = (encoding + image[y, start_x:end_x+1, 0]) / 2
# new_image[y, start_x:end_x+1, 2] = color
flaging = False
start_x = 0
for x in range(W):
for y in range(H-1):
if np.abs((image[y, x] - image[y+1, x])).sum() < delta:
if not flaging:
flaging = True
start_y = y
else:
if flaging:
flaging = False
end_y = y
if (end_y - start_y) > y_length:
encoding = np.linspace(0, 1, end_y - start_y + 1)
# color = image[start_y:end_y+1, x, :].mean()
new_image[start_y:end_y+1, x, 1] = (encoding + image[start_y:end_y+1, x, 1]) / 2
# new_image[start_y:end_y+1, x, 2] = (new_image[start_y:end_y+1, x, 2] + color) / 2
flaging = False
start_y = 0
return new_image