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
|
"""
$python xmlTotxt.py -n $FOLDER -u $LABELME_USER
--classes $CLASS_1 $CLASS_2 $CLASS_3...
FOLDER: LabelMe project folder name
LABELME_USER: LabelMe user name
CLASS_i: Class labels defined on LabelMe
"""
import argparse
import logging
import os
import xml.etree.ElementTree as ET
from os.path import join
from shutil import copyfile
def convert(size, in_x, in_y):
dw = 1./size[0]
dh = 1./size[1]
x = (in_x[0] + in_x[1])/2.0
y = (in_y[0] + in_y[1])/2.0
w = in_x[1] - in_x[0]
h = in_y[1] - in_y[0]
x = x*dw
w = w*dw
y = y*dh
h = h*dh
return (x, y, w, h)
def convert_annotation(in_dir, out_dir, image_id, out_id, classes):
in_file = open("%s/%s.xml" % (in_dir, image_id))
o_file = open(out_dir + "/%s.txt" % out_id, "w")
tree = ET.parse(in_file)
root = tree.getroot()
size = root.find("imagesize")
h = float(size.find("nrows").text)
w = float(size.find("ncols").text)
for obj in root.iter("object"):
X = []
Y = []
cls = obj.find("name").text
if cls not in classes:
logging.debug("%s is not in the selected class" % cls)
continue
cls_id = classes.index(cls)
for pt in obj.find("polygon").findall("pt"):
X.append(float(pt.find("x").text))
Y.append(float(pt.find("y").text))
if (len(X) < 2 or len(Y) < 2):
logging.warning("%s doesn't have sufficient info, ignore" % cls)
continue
X = list(set(X))
X.sort()
Y = list(set(Y))
Y.sort()
bb = convert((w, h), X, Y)
o_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + "\n")
def find_output_id(out_img_dir, image_id, suffix):
counter = 0
out_img_path = join(out_img_dir, image_id + "." + suffix)
out_id = image_id
while os.path.exists(out_img_path):
logging.info("%s exists" % out_img_path)
counter += 1
out_id = image_id + "_" + str(counter)
out_img_path = join(out_img_dir, out_id + "." + suffix)
return out_id, out_img_path
def main():
parser = argparse.ArgumentParser(
description="Simple tool to make the scene image better."
)
parser.add_argument(
"-v", "--verbosity", action="count",
help="increase output verbosity"
)
parser.add_argument(
"-r", "--root", type=str, default=None,
help="Specify the root directory (default: PWD)"
)
parser.add_argument(
"-n", "--name", type=str, default="youtube_09",
help="Specify the name of the original video (default: youtube_09)"
)
parser.add_argument(
"-u", "--user", type=str, default="V",
help="Specify the username (default: V)"
)
parser.add_argument(
"-o", "--outdir", type=str, default="test2017",
help="Output dir in root, must end with 2017 (default: test2017)"
)
parser.add_argument(
"--classes", nargs="+", type=str, default=["fighting", "dog"],
help="Classes to be trained. Default: [fighting, dog]"
)
parser.add_argument(
"-d", "--delete", action="store_true",
help="Use this option to clean up train.txt"
)
args = parser.parse_args()
log_level = logging.WARNING
if args.verbosity == 1:
log_level = logging.INFO
elif args.verbosity >= 2:
log_level = logging.DEBUG
logging.basicConfig(level=log_level,
format="[xmlTotxt: %(levelname)s] %(message)s")
logging.info(args.classes)
if args.root is None:
args.root = os.getcwd()
in_dir = join(args.root, "Annotations", "users",
args.user, args.name)
in_img_dir = join(args.root, "Images", "users",
args.user, args.name)
out_dir = join(args.root, args.outdir, "labels")
out_img_dir = join(args.root, args.outdir, "JPEGImages")
if not os.path.exists(out_dir):
os.makedirs(out_dir)
if not os.path.exists(out_img_dir):
os.makedirs(out_img_dir)
foutput = join(args.root, "train.txt")
if args.delete:
output = open(foutput, "w")
else:
output = open(foutput, "a")
for _img_path in os.listdir(in_img_dir):
img_path = os.path.join(in_img_dir, _img_path)
suffix = os.path.basename(_img_path).split(".")[1]
logging.debug("Find %s" % img_path)
image_id = os.path.basename(_img_path).split(".")[0]
out_id, out_img_path = find_output_id(out_img_dir, image_id, suffix)
convert_annotation(in_dir, out_dir, image_id, out_id, args.classes)
logging.info("Copy %s to %s" % (img_path, out_img_path))
copyfile(img_path, out_img_path)
output.write(out_img_path + "\n")
if __name__ == '__main__':
main()
|
