-
Notifications
You must be signed in to change notification settings - Fork 23
/
convert_CitySpaces_yolo_label.py
120 lines (95 loc) · 5.72 KB
/
convert_CitySpaces_yolo_label.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
import os
import shutil
import random
import cv2
import numpy as np
import json
from tqdm import tqdm
total_class_names_set = set()
def polygon2bbox(polygon, h, w):
xmin, xmax, ymin, ymax = w, 0, h, 0
for [ptx, pty] in polygon:
xmin, xmax = min(xmin, ptx), max(xmax, ptx)
ymin, ymax = min(ymin, pty), max(ymax, pty)
return [xmin, xmax, ymin, ymax]
def convert_box(size, box):
dw, dh = 1. / size[0], 1. / size[1]
x, y, w, h = (box[0] + box[1]) / 2.0 - 1, (box[2] + box[3]) / 2.0 - 1, box[1] - box[0], box[3] - box[2]
return x * dw, y * dh, w * dw, h * dh
def convert_label(anno_abs_path, lb_path, image_id, class_names):
if not os.path.exists(anno_abs_path):
return image_id + "_None"
in_file = json.load(open(anno_abs_path))
out_file = open(lb_path, 'w')
h, w = in_file["imgHeight"], in_file["imgWidth"] # 1024, 2048
label_dict_list = in_file["objects"]
is_nolabel_flag = True
for i, label_dict in enumerate(label_dict_list):
label = label_dict["label"]
total_class_names_set.add(label)
if label in class_names:
is_nolabel_flag = False
polygon = label_dict["polygon"]
bbox = polygon2bbox(polygon, h, w) # 'xmin', 'xmax', 'ymin', 'ymax'
bb = convert_box((w, h), bbox)
cls_id = class_names.index(label) # class id
out_file.write(" ".join([str(a) for a in (cls_id, *bb)]) + '\n')
if is_nolabel_flag:
return image_id + "_Nolabel"
else:
return None
if __name__ == "__main__":
'''
https://www.cityscapes-dataset.com/downloads/
https://github.com/mcordts/cityscapesScripts
'''
class_names = [ 'bus', 'bicycle', 'car', 'motorcycle', 'person', 'rider', 'train', 'truck' ] # number of selected classes, nc=8
data_root_path = "/datasdc/zhouhuayi/dataset/domain_adaptation/CityScapes" # train 2975, val 500
if os.path.exists(os.path.join(data_root_path, "yolov5_format")):
shutil.rmtree(os.path.join(data_root_path, "yolov5_format"))
os.mkdir(os.path.join(data_root_path, "yolov5_format"))
for image_set in ["train", "val"]:
imgs_path = os.path.join(data_root_path, "yolov5_format", "images", image_set)
if not os.path.exists(os.path.join(data_root_path, "yolov5_format", "images")):
os.mkdir(os.path.join(data_root_path, "yolov5_format", "images"))
if not os.path.exists(imgs_path):
os.mkdir(imgs_path)
lbs_path = os.path.join(data_root_path, "yolov5_format", "labels", image_set)
if not os.path.exists(os.path.join(data_root_path, "yolov5_format", "labels")):
os.mkdir(os.path.join(data_root_path, "yolov5_format", "labels"))
if not os.path.exists(lbs_path):
os.mkdir(lbs_path)
ori_imgs_path = os.path.join(data_root_path, "leftImg8bit", image_set)
ori_anno_path = os.path.join(data_root_path, "gtFine", image_set)
city_names = os.listdir(ori_imgs_path)
error_list = []
for city_name in tqdm(city_names):
ori_imgs_path_city = os.path.join(ori_imgs_path, city_name)
ori_anno_path_city = os.path.join(ori_anno_path, city_name)
for img_name in os.listdir(ori_imgs_path_city):
img_abs_path = os.path.join(ori_imgs_path_city, img_name) # old img path
id = img_name.replace("_leftImg8bit.png", "")
anno_abs_path = os.path.join(ori_anno_path_city, id+"_gtFine_polygons.json")
lb_path = os.path.join(lbs_path, id+".txt") # new label path
res = convert_label(anno_abs_path, lb_path, id, class_names) # convert labels to YOLO format
if os.path.exists(lb_path):
# shutil.copy(img_abs_path, os.path.join(imgs_path, id+".jpg")) # move image
os.system("ln -s %s %s"%(img_abs_path, os.path.join(imgs_path, id+".jpg"))) # soft link of image
if res is not None:
error_list.append(res)
# finished one city
print(image_set, "--> error_list:", len(error_list), "\n", error_list)
print("[OK] finished one dataset %s"%(data_root_path))
print("selected class names in CityScapes are: \n %s"%(class_names))
print("all class names in CityScapes are: \n %s"%(total_class_names_set))
'''
train --> error_list: 10
['monchengladbach_000000_015561_Nolabel', 'weimar_000067_000019_Nolabel', 'weimar_000097_000019_Nolabel', 'dusseldorf_000106_000019_Nolabel', 'dusseldorf_000101_000019_Nolabel', 'bochum_000000_031152_Nolabel', 'strasbourg_000000_023854_Nolabel', 'strasbourg_000000_012934_Nolabel', 'strasbourg_000000_036016_Nolabel', 'strasbourg_000000_035571_Nolabel']
val --> error_list: 8
['lindau_000045_000019_Nolabel', 'lindau_000040_000019_Nolabel', 'lindau_000019_000019_Nolabel', 'lindau_000017_000019_Nolabel', 'lindau_000032_000019_Nolabel', 'lindau_000018_000019_Nolabel', 'lindau_000049_000019_Nolabel', 'lindau_000021_000019_Nolabel']
[OK] finished one dataset /datasdc/zhouhuayi/dataset/domain_adaptation/CityScapes
selected class names in CityScapes are:
['bus', 'bicycle', 'car', 'motorcycle', 'person', 'rider', 'train', 'truck']
all class names in CityScapes are:
{'train', 'wall', 'pole', 'rail track', 'out of roi', 'rider', 'vegetation', 'dynamic', 'trailer', 'motorcycle', 'polegroup', 'static', 'person', 'guard rail', 'bicycle', 'ridergroup', 'license plate', 'bridge', 'traffic light', 'bicyclegroup', 'building', 'cargroup', 'road', 'ground', 'car', 'truckgroup', 'rectification border', 'ego vehicle', 'sky', 'motorcyclegroup', 'fence', 'sidewalk', 'parking', 'tunnel', 'traffic sign', 'persongroup', 'bus', 'terrain', 'caravan', 'truck'}
'''