-
Notifications
You must be signed in to change notification settings - Fork 0
/
yolo_object_cutter.py
174 lines (150 loc) · 6.26 KB
/
yolo_object_cutter.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
# -*- coding: utf-8 -*-
# Author: Wei Jia
# Project: yolov5
# File: yolo_object_cutter.py
"""
同时支持Pillow和OpenCV进行裁剪操作(优先Pillow)
适用于对于以此目录结构存储的数据集(yolov5官方的coco128数据集目录结构):
image_dataset
- images
- no1.jpg
- no2.jpg
...
- labels
- no1.txt
- no2.txt
...
使用此程序可以将数据集中每张图片标注的内容对应的小图裁剪下来并保存到目标文件夹
目标文件夹的目录结构为
dst_dir
- no1
- no1_0_x1_y1_x2_y2.jpg
- no1_1_x1_y1_x2_y2.jpg
...
- no2
- no2_0_x1_y1_x2_y2.jpg
...
...
(PS:小图命名规则为{文件名_类别索引_(目标框)左上角x坐标_左上角y坐标_右下角x坐标_右下角y坐标.图像扩展名})
"""
from pathlib import Path
import argparse
from concurrent.futures import ProcessPoolExecutor, as_completed
NO_PILLOW = False
try:
from PIL import Image, ImageOps
except ModuleNotFoundError:
NO_PILLOW = True
import cv2
def split_list(lst, n):
if n <= 0:
raise ValueError("The number of splits must be greater than 0")
if n > len(lst):
raise ValueError("The number of splits must be less than or equal to the length of the list")
# Calculate the size of each chunk
chunk_size = len(lst) // n
remainder = len(lst) % n
# Create the chunks
chunks = []
start = 0
for i in range(n):
end = start + chunk_size + (1 if i < remainder else 0)
chunks.append(lst[start:end])
start = end
return chunks
class YoloDatasetObjectCutter:
def __init__(self, dataset_path, save_path):
self.dataset_path = Path(dataset_path)
if not self.dataset_path.exists() or not self.dataset_path.is_dir():
raise ValueError(f"{dataset_path} is not exist or not a dir.")
self._check_dataset()
self.save_path = Path(save_path)
if not self.save_path.exists():
self.save_path.mkdir(parents=False)
print(f"数据集路径:{self.dataset_path.absolute()}\n存储路径:{self.save_path.absolute()}")
self.dataset_meta = self._get_dataset_meta()
def _check_dataset(self):
if not self.dataset_path.joinpath('images').exists():
raise ValueError(f"{self.dataset_path.joinpath('images')} is not exist.")
if not self.dataset_path.joinpath('labels').exists():
raise ValueError(f"{self.dataset_path.joinpath('labels')} is not exist.")
def _get_dataset_meta(self):
meta = {}
image_files = list(self.dataset_path.joinpath('images').iterdir())
label_files_stem = [i.stem for i in self.dataset_path.joinpath('labels').iterdir() if i.suffix == '.txt']
for i in image_files:
if i.stem in label_files_stem:
meta[i.stem] = (i, self.dataset_path.joinpath('labels', f'{i.stem}.txt'))
return meta
def save_result(self, image: Path, label: Path):
cuts = self.cut_single(image, label)
if cuts:
save_cuts_path = self.save_path.joinpath(image.stem)
save_cuts_path.mkdir(exist_ok=True)
for c in cuts:
file_name = f"{image.stem}_{'_'.join(list(map(lambda x: str(x), c[:5])))}{image.suffix}"
if not NO_PILLOW:
c[-1].save(save_cuts_path.joinpath(file_name))
else:
cv2.imwrite(str(save_cuts_path.joinpath(file_name)), c[-1])
def save_result_batch(self, batch: list[tuple[Path, Path]], log):
total = len(batch)
processed = 0
for i in batch:
self.save_result(i[0], i[1])
processed += 1
if log:
print(f"\r总共: {total}个, 已处理: {processed}个, "
f"已完成 {round((processed / total) * 100, 2)}%", end='')
def run(self, workers: int = 1):
if workers == 1:
self.save_result_batch(list(self.dataset_meta.values()), log=True)
elif workers > 1:
chunks = split_list(list(self.dataset_meta.values()), workers)
print(f"总共{workers}个进程正在进行处理,请稍后...")
with ProcessPoolExecutor(max_workers=workers) as executor:
futures = {executor.submit(self.save_result_batch, chuk, log=False) for chuk in chunks}
completed = 0
for future in as_completed(futures):
completed += 1
print(f"总共{workers}个进程处理, 进程:{id(future)}完成, "
f"已完成{completed}个进程")
else:
raise ValueError('invalid workers.')
@classmethod
def cut_single(cls, image: Path, label: Path):
if not NO_PILLOW:
image = Image.open(image)
image = ImageOps.exif_transpose(image)
imw, imh = image.size
else:
image = cv2.imread(str(image))
imh, imw = image.shape[:2]
with label.open('r', encoding='utf-8') as f:
label = f.read().strip().splitlines()
cuts = []
for line in label:
boxes = line.strip().split(' ')
cls_idx = int(boxes[0])
centerx, centery, w, h = map(lambda x: float(x), boxes[1:])
centerx, w = centerx * imw, w * imw
centery, h = centery * imh, h * imh
x1, y1 = round(centerx - (w / 2)), round(centery - (h / 2))
x2, y2 = round(centerx + (w / 2)), round(centery + (h / 2))
if not NO_PILLOW:
cut = image.crop((x1, y1, x2, y2))
else:
cut = image[y1:y2, x1:x2]
cuts.append((cls_idx, x1, y1, x2, y2, cut))
return cuts
def main(_args):
cutter = YoloDatasetObjectCutter(_args.dataset_path, _args.save_path)
cutter.run(workers=_args.workers)
print("处理完毕")
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='yolo dataset object cutter')
parser.add_argument('-src', '--dataset-path', type=str, help='dataset path')
parser.add_argument('-dst', '--save-path', type=str, help='cut image save path')
parser.add_argument('-w', '--workers', default=1, type=int, help='multiprocess workers')
args = parser.parse_args()
main(args)