forked from annie7777/Data_Converter
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathconvert_xml2yolo_per_quad.py
118 lines (88 loc) · 3.26 KB
/
convert_xml2yolo_per_quad.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
import json
import os
import glob
import pandas as pd
import argparse
import xml.etree.ElementTree as ET
import cv2
import numpy as np
from skimage.measure import label, regionprops
import csv
label_dir = 'quads'
datasets = glob.glob(os.path.join(label_dir, '*'))
print(datasets)
if not os.path.exists('images'):
os.mkdir('images')
if not os.path.exists('labels'):
os.mkdir('labels')
classnames = ['ST','GT', 'HG', 'TC', 'Pink', 'BB','KB', 'FB', 'PF', 'CPF']
overlay_flag = False
def write_txt(file, data):
with open(file, 'a') as f:
writer = csv.writer(f, delimiter = ' ')
writer.writerow(data)
for dataset in datasets:
xml_files = glob.glob(os.path.join(dataset, '*.xml'))
if True:#dataset=='labelled/310_16B_GPS':
for xml_file in xml_files:
img_path = xml_file.replace('xml', 'jpg')
mask_path = img_path.replace('quads', 'masks')
img = cv2.imread(img_path)
mask = cv2.imread(mask_path, 0)
ret, thresh = cv2.threshold(mask, 10, 255, cv2.THRESH_BINARY)
label_img = label(thresh)
regions = regionprops(label_img)[0]
minr,minc,maxr,maxc = regions.bbox
quads_images = img[minr:maxr, minc:maxc, :]
quad_h, quad_w = quads_images.shape[0], quads_images.shape[1]
new_name = os.path.basename(dataset) + '_' +os.path.basename(img_path)[:-3]+'png'
print(new_name, quad_h, quad_w)
tree = ET.parse(xml_file)
root = tree.getroot()
annotations = []
for member in root.findall('object'):
value = (root.find('filename').text,
int(root.find('size')[0].text),
int(root.find('size')[1].text), member[0].text,
int(member[4][0].text), int(member[4][1].text),
int(member[4][2].text), int(member[4][3].text))
class_name = member[0].text
class_num = classnames.index(class_name)
x_tl, y_tl, x_br, y_br = float(member[4][0].text), float(member[4][1].text),float(member[4][2].text), float(member[4][3].text)
coords = [x_tl-minc, y_tl-minr, x_br-minc, y_br-minr]
if all(i<0 for i in coords):
print(new_name)
continue
else:
#qx_tl, qy_tl, qx_br, qy_br
newcoords = []
for i in coords:
if i < 0:
i = 0
newcoords.append(i)
x_min = newcoords[0]
y_min = newcoords[1]
x_max = newcoords[2]
y_max = newcoords[3]
if x_max > quads_images.shape[1]:
x_max = quads_images.shape[1]
if y_max > quads_images.shape[0]:
y_max = quads_images.shape[0]
if x_min > quads_images.shape[1] or y_min > quads_images.shape[0]:
continue
else:
ww, hh = x_max-x_min, y_max-y_min
xc = x_min + ww/2
yc = y_min + hh/2
xc /= quads_images.shape[1]
yc /= quads_images.shape[0]
ww /= quads_images.shape[1]
hh /= quads_images.shape[0]
write_txt("labels/{}.txt".format(new_name[:-4]), [class_num, xc, yc, ww, hh]) #xc/quad_w, yc/quad_h, ww/quad_w, hh/quad_h])
if overlay_flag:
pt1 = (int(xc*quads_images.shape[1]-ww*quads_images.shape[1]/2), int(yc*quads_images.shape[0]-hh*quads_images.shape[0]/2))
pt2 = (pt1[0]+int(ww*quads_images.shape[1]), pt1[1]+int(hh*quads_images.shape[0]))
quads_images = cv2.rectangle(quads_images, pt1, pt2, (0, 255, 0), 1)
cv2.imwrite(os.path.join('images', new_name), quads_images)
else:
cv2.imwrite(os.path.join('images', new_name), quads_images)