def main(args: List[str]):
if len(args) > 1:
data_folder_str = args[1]
else:
data_folder_str = folder.data_folder()
voc_label_path = Path(__file__).parent.parent / data_folder_str / folder.annotation_folder() # VOC格式标记的数据
image_path = Path(__file__).parent.parent / data_folder_str / folder.img_folder() # 图片数据
image_sets_path = Path(__file__).parent.parent / data_folder_str / folder.split_folder() # 训练集测试集分类数据
coco_labels_path = Path(__file__).parent.parent / data_folder_str / folder.coco_labels_folder() # COCO格式标记的数据
if not coco_labels_path.exists():
coco_labels_path.mkdir()
for name in ('train', 'test'):
image_ids = \
set((image_sets_path / f'{name}.txt').open('r').read().strip().split()) # [0001 0002 ...]
xmls = []
# 给coco查找图片用的
# print(image_path.exists())
# print(image_path)
for image_xml in voc_label_path.glob("*.xml"):
# print(image)
# 图片是后缀名且文件名在对应标签里
if image_xml.stem in image_ids:
xmls.append(image_xml)
# print(xmls)
convert(xmls, coco_labels_path / f'{name}.json')
def main(args: list):
if len(args) > 1:
data_folder_str = args[1]
else:
data_folder_str = folder.data_folder()
image_path = Path(
__file__).parent.parent / data_folder_str / folder.img_folder() # 图片数据
mean, standard = compute(image_path)
print("mean", str(mean).replace(' ', ','))
print("std", str(standard).replace(' ', ','))
def main(args: list):
if len(args) > 1:
data_folder = Path(args[1])
else:
data_folder = Path(folder.data_folder())
label_folder_path = Path(
__file__).parent.parent / data_folder / folder.annotation_folder(
) # 标记的数据
image_folder_path = Path(
__file__).parent.parent / data_folder / folder.img_folder() # 图片数据
# print(label_folder_path.exists())
annotation_set = set(f.stem for f in label_folder_path.glob("*.xml"))
# print(annotation_set)
for img_path in image_folder_path.rglob(r".*\.png|.*\.jpg|.*\.jpeg"):
if img_path.stem not in annotation_set:
xml = create_xml(img_path)
# print(xml)
filename = f'{img_path.stem}.xml'
# print(filename)
with (label_folder_path / filename).open('w', newline='\n') as f:
f.write(xml)
def main(args: list):
if len(args) > 1:
data_folder_str = args[1]
else:
data_folder_str = folder.data_folder()
voc_label_path = Path(
__file__).parent.parent / data_folder_str / folder.annotation_folder(
) # VOC格式标记的数据
image_path = Path(
__file__).parent.parent / data_folder_str / folder.img_folder() # 图片数据
image_sets_path = Path(
__file__).parent.parent / data_folder_str / folder.split_folder(
) # 训练集测试集分类数据
yolo_labels_path = Path(
__file__).parent.parent / data_folder_str / folder.yolo_labels_folder(
) # YOLO格式标记的数据
if not yolo_labels_path.exists():
yolo_labels_path.mkdir()
for name in ('train', 'test'):
image_ids = \
set((image_sets_path / f'{name}.txt').open('r').read().strip().split()) # [0001 0002 ...]
list_file = (Path(__file__).parent.parent /
f'darknet_{name}.txt').open('wb')
# 给darknet查找图片用的，保存在当前目录
# print(image_path.exists())
# print(image_path)
for image in image_path.iterdir():
# print(image)
# 图片是后缀名且文件名在对应标签里
if image.suffix in ('.png', '.jpeg', '.jpg', '.gif',
'.bmp') and image.stem in image_ids:
list_file.write(bytes(image.absolute()))
list_file.write(bytes('\n', 'utf-8'))
convert_annotation(voc_label_path / f'{image.stem}.xml',
yolo_labels_path / f'{image.stem}.txt')
list_file.close()
def main(args: list):
if len(args) > 1:
ratio = float(args[1])
else:
ratio = 0.2
if len(args) > 2:
data_folder = Path(args[2])
else:
data_folder = Path(folder.data_folder())
label_path = Path(
__file__).parent.parent / data_folder / folder.annotation_folder(
) # 标记的数据
name_path = Path(
__file__).parent.parent / data_folder / folder.split_folder(
) # 用于训练和测试样本的名字文件的路径
list_path = list(f'{f.stem}\n' for f in label_path.glob("*.xml"))
train_list, test_list = split(list_path, True, ratio)
print(train_list)
print(test_list)
for name in ('train', 'test'):
pth = name_path / f"{name}.txt"
with pth.open('w', newline='\n') as f:
f.writelines(eval(f"{name}_list"))