def _add_images_doc(self, path: str, import_path: str, url_prefix: str):
"""
为文件夹下的图片添加信息字典文件(id为不含扩展的文件名)
@param {str} path - 要处理的文件目录
"""
_import_path = os.path.realpath(import_path)
# 处理当前文件夹
_file_list = FileTool.get_filelist(path,
regex_str=r'^((?!\.json$).)*$',
is_fullname=True)
for _file in _file_list:
_ext = FileTool.get_file_ext(_file)
_json_file = _file[0:-len(_ext)] + 'json'
if os.path.exists(_json_file):
# 字典文件已存在，无需处理
continue
# 生成并写入字典
_file_name = FileTool.get_file_name_no_ext(_file)
_url = os.path.realpath(_file)[len(_import_path):].replace(
'\\', '/').lstrip('/')
_url = '%s/%s' % (url_prefix, _url)
_image_doc = {'id': _file_name, 'url': _url, 'path': _file}
_json_str = json.dumps(_image_doc, ensure_ascii=False)
with open(_json_file, 'wb') as _fid:
_fid.write(_json_str.encode(encoding='utf-8'))
# 处理子文件夹
_sub_dir_list = FileTool.get_dirlist(path)
for _sub_dir in _sub_dir_list:
self._add_images_doc(_sub_dir, import_path, url_prefix)
def rename_file_to_num(cls, path: str, start_index: int = 1) -> int:
"""
重命名文件为数字序号
@param {str} path - 要处理的文件夹
@param {int} start_index=1 - 开始序号
@returns {int} - 返回当前序号
"""
# 处理当前目录
_start_index = start_index
_path = os.path.realpath(path)
_file_list = FileTool.get_filelist(_path, is_fullname=False)
for _file in _file_list:
_ext = FileTool.get_file_ext(_file)
os.rename(
os.path.join(_path, _file),
os.path.join(
_path,
StringTool.fill_fix_string(str(_start_index), 10, '0') +
'.' + _ext))
_start_index += 1
# 处理子目录
_sub_dir_list = FileTool.get_dirlist(_path)
for _sub_dir in _sub_dir_list:
_start_index = cls.rename_file_to_num(_sub_dir,
start_index=_start_index)
# 返回当前序号值
return _start_index
def labelimg_rename_filename(cls, path: str, fix_len: int = 10):
"""
重名名labelimg对应目录下的文件名（图片文件和标注文件同步修改）
@param {str} path - 要修改文件名的路径
@param {int} fix_len=10 - 文件名长度
"""
_path = os.path.realpath(path)
_files = FileTool.get_filelist(path=_path, is_fullname=False)
_index = 1
for _file in _files:
_file_ext = FileTool.get_file_ext(_file)
if _file_ext == 'xml':
# 标签文件不处理
continue
_file_no_ext = FileTool.get_file_name_no_ext(_file)
# 获取最新的文件名
while True:
_new_name = StringTool.fill_fix_string(str(_index),
fix_len,
'0',
left=True)
_new_file = _new_name + '.' + _file_ext
_index += 1
if os.path.exists(os.path.join(path, _new_file)):
# 文件名已存在
_index += 1
continue
# 文件名不存在，跳出循环
break
# 修改文件名
os.rename(os.path.join(_path, _file),
os.path.join(_path, _new_file))
if os.path.exists(os.path.join(_path, _file_no_ext + '.xml')):
# 需要修改标签文件
_xml_file = _new_name + '.xml'
os.rename(os.path.join(_path, _file_no_ext + '.xml'),
os.path.join(_path, _xml_file))
# 修改标签文件内容
_tree = ET.parse(os.path.join(_path, _xml_file))
_root = _tree.getroot()
_root.find('filename').text = _new_file
_root.find('path').text = os.path.join(_path, _new_file)
_tree.write(os.path.join(_path, _xml_file),
encoding='utf-8',
method="xml",
xml_declaration=None)
def load_predef_by_file(self, file: str, encoding: str = 'utf-8'):
"""
通过文件装载预定义模板
@param {str} file - 要装载的文件名
@param {str} encoding='utf-8' - 文件编码
"""
_ext = FileTool.get_file_ext(file).lower()
if _ext == 'json':
with open(file, 'r', encoding=encoding) as _fp:
_json = _fp.read()
_config = json.loads(_json)
else:
raise AttributeError('Not support file type [%s]!' % _ext)
# 装载到机器人
self.load_predef_by_config(_config)
def import_images(self, path: str, pipeline: str, encoding: str = 'utf-8'):
"""
将指定路径的图片导入搜索库
@param {str} path - 图片及信息字典所在路径，包含文件:
图片文件，例如"abc.jpg"
对应的json信息字典文件，例如"abc.json"，文件内容按标准json字符串格式编写
注：json中可以通过添加collection域指定该图片的所属分类集合名
@param {str} pipeline - 处理管道标识
@param {str} encoding='utf-8' - json文件的编码
"""
_pipeline_obj = self._get_pipeline(pipeline)
_file_list = FileTool.get_filelist(path,
regex_str=r'^((?!\.json$).)*$',
is_fullname=True)
for _file in _file_list:
try:
# 获取图片的信息字典
_ext = FileTool.get_file_ext(_file)
_json_file = _file[0:-len(_ext)] + 'json'
if not os.path.exists(_json_file):
self.log_debug(
'Json file not exists, not imported: [%s]!' % _file)
continue
with open(_json_file, 'r', encoding=encoding) as _fid:
_image_doc = json.loads(_fid.read())
# 导入图片
_collection = _image_doc.get('collection', '')
with open(_file, 'rb') as _fid:
self._image_to_search_db(_fid.read(),
_image_doc,
_pipeline_obj,
init_collection=_collection)
# 输出日志
self.log_debug('image [%s] imported success' % _file)
except:
self.log_debug('image [%s] import error: %s' %
(_file, traceback.format_exc()))
def _get_pic_file_list(cls, input_path: str) -> list:
"""
获取制定目录下的所有图片文件清单
@param {str} input_path - 要处理的目录
@returns {list} - 文件清单列表
"""
_list = []
# 先获取当前目录下的所有xml文件
for _file in FileTool.get_filelist(input_path, is_fullname=True):
_ext = FileTool.get_file_ext(_file)
if _ext.lower() in ('jpg', 'jpeg'):
_list.append(_file)
# 获取子目录
for _dir in FileTool.get_dirlist(input_path):
_temp_list = cls._get_pic_file_list(_dir)
_list.extend(_temp_list)
return _list
def labelimg_del_not_rgb_pic(cls, path: str):
"""
删除位深不为RGB三通道的图片
（解决image_size must contain 3 elements[4]报错）
@param {str} path - 要处理的路径
"""
_path = os.path.realpath(path)
# 遍历所有子目录
_sub_dirs = FileTool.get_dirlist(path=_path, is_fullpath=True)
for _dir in _sub_dirs:
# 递归删除子目录的信息
cls.labelimg_del_not_rgb_pic(_dir)
# 检查自己目录下的图片
_files = FileTool.get_filelist(path=_path, is_fullname=False)
for _file in _files:
_file_ext = FileTool.get_file_ext(_file)
if _file_ext == 'xml':
# 标签文件不处理
continue
# 打开图片判断位深
_fp = open(os.path.join(_path, _file), 'rb')
_img = Image.open(_fp)
if _img.mode != 'RGB':
# 需要删除掉
_fp.close()
_img_file = os.path.join(_path, _file)
_xml_file = os.path.join(
_path,
FileTool.get_file_name_no_ext(_file) + '.xml')
print('delete %s' % _img_file)
FileTool.remove_file(_img_file)
if os.path.exists(_xml_file):
FileTool.remove_file(_xml_file)
else:
_fp.close()
def UploadFile(cls,
upload_type: str,
note: str,
interface_seq_id: str,
methods=['POST']):
"""
上传文件(单文件上传) (/api/Qa/UploadFiles/<upload_type>/<note>/<interface_seq_id>)
@param {str} upload_type - 文件类型，必须在UploadFileConfig表中有配置
@param {str} note - 文件注解
@param {str} interface_seq_id - 客户端序号，客户端可传入该值来支持异步调用
@return {str} - 返回回答的json字符串
status : 处理状态
00000 - 成功, 返回一条回答
10001 - 没有指定上传文件
2XXXX - 处理失败
msg : 处理状态对应的描述
answer_type: 'text'或'json'，指示返回的答案是文本数组，还是一个json对象
answers : 回答内容
url : 文件上传后的url，含文件名和url路径
"""
_ret_json = {
'interface_seq_id': interface_seq_id,
'status': '00000',
'msg': 'success',
'answer_type': 'text',
'answers': [],
'url': ''
}
_qa_loader = RunTool.get_global_var('QA_LOADER')
try:
if 'file' not in request.files or request.files[
'file'].filename == '':
_ret_json['status'] = '10001'
_ret_json['msg'] = 'No file upload!'
return jsonify(_ret_json)
# 获取上传类型配置
_upload_config = UploadFileConfig.get_or_none(
UploadFileConfig.upload_type == upload_type)
if _upload_config is None:
_ret_json['status'] = '10002'
_ret_json['msg'] = 'upload type not exists!'
return jsonify(_ret_json)
# 检查文件大小
if _upload_config.size > 0:
if request.content_length > _upload_config.size * 1024 * 1024:
_ret_json['status'] = '10003'
_ret_json['msg'] = 'upload file size to large!'
return jsonify(_ret_json)
# 检查文件类型是否支持
_file = request.files['file']
_old_filename = _file.filename
_file_ext = FileTool.get_file_ext(_old_filename)
_allow_ext = eval(_upload_config.exts.upper())
if len(_allow_ext) > 0 and _file_ext.upper() not in _allow_ext:
_ret_json['status'] = '10004'
_ret_json['msg'] = 'Type [%s] not allow upload [.%s] file!' % (
upload_type, _file_ext)
return jsonify(_ret_json)
# 处理新的文件名
def _replace_var_fun(m):
_match_str = m.group(0)
_value = None
if _match_str.startswith('{$datetime='):
# 按格式化字符替换当前的时间
_key = _match_str[11:-2]
_value = datetime.datetime.now().strftime(_key)
elif _match_str.startswith('{$uuid='):
# 指定uuid字符类型
_key = _match_str[7:-2]
str(uuid.uuid1())
_value = eval('str(uuid.uuid%s())' % _key)
elif _match_str.startswith('{$random'):
# 产生指定两个整数之间的随机数，总位数与最大的数一致，左补零
_key = _match_str[8:-2]
_args = eval('(%s)' % _key)
_value = StringTool.fill_fix_string(
str(random.randint(*_args)), len(_args[1]), '0')
elif _match_str.startswith('{$file_ext='):
# 原文件扩展名
_value = _file_ext
elif _match_str.startswith('{$file_name='):
# 原文件指定位置的字符
_key = _match_str[12:-2]
_args = eval('(%s)' % _key)
if len(_args) > 1:
_value = _old_filename[_args[0]:_args[1]]
else:
_value = _old_filename[_args[0]:]
if _value is not None:
return str(_value)
else:
return _match_str
if _upload_config.rename == '':
_new_filename = _old_filename
else:
_new_filename = re.sub(r'\{\$.+?\$\}', _replace_var_fun,
_upload_config.rename, re.M)
# 处理保存文件路径和url路径
if _upload_config.url != '':
_ret_json['url'] = '%s/%s' % (_upload_config.url,
_new_filename)
_save_path = os.path.realpath(
os.path.join(_qa_loader.execute_path, _upload_config.save_path,
_new_filename))
# 创建文件目录
FileTool.create_dir(os.path.split(_save_path)[0], exist_ok=True)
# 保存文件
_file.save(_save_path)
# 上传后处理
_after = eval(_upload_config.after)
if len(_after) > 0:
_after_fun = _qa_loader.plugins['upload_after'][_after[0]][
_after[1]]
_status, _msg, _answers = _after_fun(upload_type, note,
_new_filename, _save_path,
_ret_json['url'],
**_after[2])
_ret_json['status'] = _status
_ret_json['msg'] = _msg
if len(_answers) > 0 and type(_answers[0]) == dict:
_ret_json['answer_type'] = 'json'
_ret_json['answers'] = _answers[0]
else:
_ret_json['answers'] = _answers
if _ret_json['status'] != '00000':
# 后处理失败，删除文件
FileTool.remove_file(_save_path)
if _qa_loader.logger:
_qa_loader.logger.debug(
'remove upload file [dest:%s][source:%s] when after deal error[%s]: %s'
% (_new_filename, _old_filename, _status, _msg))
except:
if _qa_loader.logger:
_qa_loader.logger.error('Exception: %s' %
traceback.format_exc(),
extra={'callFunLevel': 1})
_ret_json = {
'interface_seq_id': interface_seq_id,
'status': '20001',
'msg': '上传文件异常'
}
return jsonify(_ret_json)
def labelimg_crop_pic_by_flags(cls,
path: str,
dest_path: str,
copy_no_flag_pic: bool = True,
with_sub_dir: bool = True,
fix_len: int = 10):
"""
根据标注进行图片截图处理
@param {str} path - 需要处理的目录
@param {str} dest_path - 截图图片存放目录
@param {bool} copy_no_flag_pic=True - 直接复制没有标注的图片
@param {bool} with_sub_dir=True - 是否按原目录结构存储图片
@param {int} fix_len=10 - 图片重命名的文件名长度
@returns {iter_list} - 通过yield返回的处理进度信息清单
[总文件数int, 当前已处理文件数int, 是否成功]
"""
try:
# 获取所有要处理的图片清单
_file_list = cls._get_pic_file_list(path)
_total = len(_file_list)
_deal_num = 0
# 先返回进度情况
if _total == 0:
yield [_deal_num, _total, True]
return
# 创建复制文件夹
FileTool.create_dir(dest_path, exist_ok=True)
# 遍历处理
_rename_index = 1
_src_path = os.path.realpath(path)
_dest_path = os.path.realpath(dest_path)
for _file in _file_list:
# 当前进展
yield [_deal_num, _total, True]
# 路径处理
_file_path, _filename = os.path.split(_file)
if with_sub_dir:
# 创建子目录
_dest_path = os.path.join(
os.path.realpath(dest_path),
os.path.realpath(_file_path)[len(_src_path):].strip(
'/\\'))
FileTool.create_dir(_dest_path, exist_ok=True)
# 获取标注文件
_ext = FileTool.get_file_ext(_filename)
_xml_file = os.path.join(_file_path,
_filename[0:-len(_ext)] + 'xml')
if not os.path.exists(_xml_file):
# 标注文件不存在
if copy_no_flag_pic:
# 直接复制文件
shutil.copy(
_file,
os.path.join(
_dest_path,
StringTool.fill_fix_string(
str(_rename_index), fix_len, '0') + '.' +
_ext))
_rename_index += 1
# 下一个
_deal_num += 1
continue
# 将图片放入内存
with open(_file, 'rb') as _fid:
_file_bytes = _fid.read()
_image = Image.open(BytesIO(_file_bytes))
# 处理标注
_tree = ET.parse(_xml_file)
_root = _tree.getroot()
for _member in _root.findall('object'):
# 逐个标注进行处理
_crop_image = _image.crop(
(int(_member[4][0].text), int(_member[4][1].text),
int(_member[4][2].text), int(_member[4][3].text)))
_crop_image.save(os.path.join(
_dest_path,
StringTool.fill_fix_string(str(_rename_index), fix_len,
'0') + '.' + _ext),
format='JPEG')
_rename_index += 1
# 下一个
_deal_num += 1
# 返回结果
yield [_total, _total, True]
except:
print('labelimg_crop_pic_by_flags error: %s\r\n%s' %
(path, traceback.format_exc()))
yield [-1, -1, False]
def labelimg_pic_deal(cls, path: str):
"""
TFRecord图片兼容处理
1.删除位深不为RGB三通道的图片
（解决image_size must contain 3 elements[4]报错）
2.转换图片格式为jpg
3.检查xml文件的文件名和路径是否正确
@param {str} path - 要处理的路径
"""
_path = os.path.realpath(path)
# 遍历所有子目录
_sub_dirs = FileTool.get_dirlist(path=_path, is_fullpath=True)
for _dir in _sub_dirs:
# 递归删除子目录的信息
cls.labelimg_pic_deal(_dir)
# 检查自己目录下的图片
_files = FileTool.get_filelist(path=_path, is_fullname=False)
for _file in _files:
_file_ext = FileTool.get_file_ext(_file)
if _file_ext == 'xml':
# 标签文件不处理
continue
_img_file = os.path.join(_path, _file)
_file_no_ext = FileTool.get_file_name_no_ext(_file)
if _file_ext in ('png', 'gif'):
# 转换图片格式
_fp = open(_img_file, 'rb')
_img = Image.open(_fp)
_rgb_im = _img.convert('RGB')
_rgb_im.save(os.path.join(_path, _file_no_ext + '.jpg'))
_fp.close()
# 删除原文件，修改xml中的文件名
FileTool.remove_file(_img_file)
_xml_file = os.path.join(_path, _file_no_ext + '.xml')
if os.path.exists(_xml_file):
_tree = ET.parse(os.path.join(_path, _xml_file))
_root = _tree.getroot()
_root.find('filename').text = _file_no_ext + '.jpg'
_root.find('path').text = os.path.join(
_path, _file_no_ext + '.jpg')
_tree.write(os.path.join(_path, _xml_file),
encoding='utf-8',
method="xml",
xml_declaration=None)
# 修改文件名变量
_img_file = os.path.join(_path, _file_no_ext + '.jpg')
# 打开图片判断位深
_fp = open(_img_file, 'rb')
_img = Image.open(_fp)
if _img.mode != 'RGB':
# 需要删除掉
_fp.close()
_xml_file = os.path.join(
_path,
FileTool.get_file_name_no_ext(_file) + '.xml')
print('delete %s' % _img_file)
FileTool.remove_file(_img_file)
if os.path.exists(_xml_file):
FileTool.remove_file(_xml_file)
else:
_fp.close()
# 检查xml文件
_xml_file = os.path.join(_path, _file_no_ext + '.xml')
if os.path.exists(_xml_file):
_tree = ET.parse(os.path.join(_path, _xml_file))
_root = _tree.getroot()
if _root.find('filename'
).text != _file_no_ext + '.jpg' or os.path.split(
_root.find('path').text)[0] != _path:
_root.find('filename').text = _file_no_ext + '.jpg'
_root.find('path').text = os.path.join(
_path, _file_no_ext + '.jpg')
_tree.write(os.path.join(_path, _xml_file),
encoding='utf-8',
method="xml",
xml_declaration=None)
def labelimg_copy_flags_pics(cls,
input_path: str,
output_path: str,
use_mapping: bool = False,
mapping: dict = None):
"""
按类别复制图片和标注文件到指定目录
@param {str} input_path - 图片路径
@param {str} output_path - 输出路径
@param {bool} use_mapping=False - 是否使用mapping处理映射
@param {dict} mapping=None - mapping.josn字典
@returns {iter_list} - 通过yield返回的处理进度信息清单
[总文件数int, 当前已处理文件数int, 是否成功]
"""
try:
# 遍历所有文件夹，获取需要处理的文件数量
_file_list = cls._get_labelimg_annotation_file_list(input_path)
_total = len(_file_list)
_deal_num = 0
# 先返回进度情况
if _total == 0:
yield [_deal_num, _total, True]
return
# 创建复制文件夹
FileTool.create_dir(output_path, exist_ok=True)
# 遍历处理
for _file in _file_list:
# 当前进展
yield [_deal_num, _total, True]
# 逐个标注文件进行处理
_tree = ET.parse(_file)
_root = _tree.getroot()
_annotations = dict()
_annotations['filename'] = _root.find('filename').text
_annotations['file_path'] = os.path.join(
os.path.split(_file)[0], _annotations['filename'])
# 获取信息字典
_info_dict = ExtendLib.get_info_dict(_annotations['file_path'],
mapping['info_key_dict'])
# 逐个标签处理
_save_class_path = '' # 要保存到的分类路径
_is_copy = False # 标注是否已复制文件
_new_xml_name = '' # 新的xml名
for _member in _root.findall('object'):
_member_class = _member[0].text
if use_mapping:
# 使用映射处理
if _member_class == mapping['set_by_info'][
'class_name']:
# 需要获取真实的信息
if mapping['set_by_info'][
'info_tag'] in _info_dict.keys():
_member_class = _info_dict[
mapping['set_by_info']['info_tag']]
# 变更分类名
_member[0].text = _member_class
# 过滤不需要的类别
if _member_class not in mapping['class_int'].keys():
_deal_num += 1
continue
# 保存分类路径
_save_class_path = os.path.join(
output_path, mapping['class'][_member_class])
else:
# 普通分类
_save_class_path = os.path.join(
output_path, _member_class)
# 复制文件
if not _is_copy:
# 处理重复文件名
_file_name = FileTool.get_file_name_no_ext(
_annotations['filename'])
_file_ext = FileTool.get_file_ext(
_annotations['filename'])
_rename_num = 1
_new_file_name = '%s.%s' % (_file_name, _file_ext)
_new_xml_name = '%s.xml' % (_file_name, )
while os.path.exists(
os.path.join(_save_class_path,
_new_file_name)):
_new_file_name = '%s_%d.%s' % (
_file_name, _rename_num, _file_ext)
_new_xml_name = '%s_%d.xml' % (_file_name,
_rename_num)
_rename_num += 1
# 创建文件夹
FileTool.create_dir(_save_class_path, exist_ok=True)
shutil.copyfile(
_annotations['file_path'],
os.path.join(_save_class_path, _new_file_name))
# 修改xml里面的文件名和文件路径
_root.find('filename').text = _new_file_name
_root.find('path').text = os.path.join(
_save_class_path, _new_file_name)
_is_copy = True
if _is_copy:
# 有修改xml内容
_tree.write(os.path.join(_save_class_path, _new_xml_name),
encoding='utf-8',
method="xml",
xml_declaration=None)
# 继续循环处理
_deal_num += 1
# 返回结果
yield [_total, _total, True]
except:
print('labelimg_copy_flags_pics error: %s\r\n%s' %
(input_path, traceback.format_exc()))
yield [-1, -1, False]
def _clean_file_path(cls, path: str, class_path: str):
"""
清理当前目录文件
@param {str} path - 要处理的目录地址
@param {str} class_path - 类目录
"""
# 处理自身目录，先获取商品信息
_info = dict()
_info_file = os.path.join(path, 'info.json')
if os.path.exists(_info_file):
with open(_info_file, 'rb') as f:
_eval = str(f.read(), encoding='utf-8')
_info = eval(_eval)
# 判断是否不处理
_shop_name = _info['店名']
# if _info['款式'] == '挂件' and _info['挂件类型'] == '':
# return
# 遍历文件进行处理
_product_num = FileTool.get_dir_name(path)
_files = FileTool.get_filelist(path)
_order = 1
for _file in _files:
_file_ext = FileTool.get_file_ext(_file).lower()
if _file_ext not in ['jpg', 'jpeg', 'png', 'bmp']:
# 不是合适的文件类型
continue
# 判断是否有括号
if _file.find('(') >= 0:
FileTool.remove_file(_file)
continue
# 判断是否匹配上要删除的图片大小
if _shop_name in DEL_SHOP_PIC_SIZE.keys() and os.path.getsize(
_file) in DEL_SHOP_PIC_SIZE[_shop_name]:
FileTool.remove_file(_file)
continue
# 修改文件名
if not FileTool.get_file_name(_file).startswith(_product_num):
os.rename(
_file,
os.path.join(
path, '%s_%s_%d.%s' %
(_product_num, 'main' if _file.find('主图') >= 0
or _file.find('main') >= 0 else 'detail', _order,
_file_ext)))
# 下一个文件
_order += 1
# 移动文件夹到指定的分类目录
_class_path = _info['款式']
if _class_path in PROP_TYPE_TRAN_DICT.keys():
_class_path = PROP_TYPE_TRAN_DICT[_info['款式']]
shutil.move(path,
os.path.join(class_path, _class_path, _product_num))
# 处理完成，返回
return
def labelimg_copy_flags_pics(cls,
input_path: str,
output_path: str,
is_cc: bool = False):
"""
按类别复制图片和标注文件到指定目录
@param {str} input_path - 图片路径
@param {str} output_path - 输出路径
@param {bool} is_cc=False - 是否CC项目
@returns {iter_list} - 通过yield返回的处理进度信息清单
[总文件数int, 当前已处理文件数int, 是否成功]
"""
try:
# 遍历所有文件夹，获取需要处理的文件数量
_file_list = cls._get_labelimg_annotation_file_list(input_path)
_total = len(_file_list)
_deal_num = 0
# 先返回进度情况
if _total == 0:
yield [_deal_num, _total, True]
return
# 创建复制文件夹
FileTool.create_dir(output_path, exist_ok=True)
# 遍历处理
for _file in _file_list:
# 当前进展
yield [_deal_num, _total, True]
# 逐个标注文件进行处理
_tree = ET.parse(_file)
_root = _tree.getroot()
_annotations = dict()
_annotations['filename'] = _root.find('filename').text
_annotations['file_path'] = os.path.join(
os.path.split(_file)[0], _annotations['filename'])
# 逐个标签处理
_save_class_path = '' # 要保存到的分类路径
_is_copy = False # 标注是否已复制文件
_is_change_class = False # 标注是否有修改分类名
_new_xml_name = '' # 新的xml名
for _member in _root.findall('object'):
_member_class = _member[0].text
if is_cc:
# CC专属的类型转换
if _member_class == '翡翠':
# 需要获取真实的信息
_info_file = os.path.join(
os.path.split(_file)[0], 'info.json')
_info = dict()
with open(_info_file, 'rb') as f:
_eval = str(f.read(), encoding='utf-8')
_info = eval(_eval)
if _info['款式'] == '挂件':
# 挂件，需要二级分类
_member_class = _info['挂件类型']
else:
# 一级分类
_member_class = _info['款式']
# 变更分类名
_member[0].text = _member_class
_is_change_class = True
# 过滤不需要的类别
if _member_class not in USE_CLASS_TEXT_LIST:
_deal_num += 1
continue
# 保存分类路径
_save_class_path = os.path.join(
output_path, cls._cc_get_class_text(_member_class))
else:
# 普通分类
_save_class_path = os.path.join(
output_path, _member_class)
# 复制文件
if not _is_copy:
# 处理重复文件名
_file_name = FileTool.get_file_name_no_ext(
_annotations['filename'])
_file_ext = FileTool.get_file_ext(
_annotations['filename'])
_rename_num = 1
_new_file_name = '%s.%s' % (_file_name, _file_ext)
_new_xml_name = '%s.xml' % (_file_name, )
while os.path.exists(
os.path.join(_save_class_path,
_new_file_name)):
_new_file_name = '%s_%d.%s' % (
_file_name, _rename_num, _file_ext)
_new_xml_name = '%s_%d.xml' % (_file_name,
_rename_num)
_rename_num += 1
# 创建文件夹
FileTool.create_dir(_save_class_path, exist_ok=True)
shutil.copyfile(
_annotations['file_path'],
os.path.join(_save_class_path, _new_file_name))
_is_copy = True
if _is_copy:
# 有复制文件
if _is_change_class:
# 有修改xml内容
_tree.write(os.path.join(_save_class_path,
_new_xml_name),
encoding='utf-8',
method="xml",
xml_declaration=None)
else:
shutil.copyfile(
_file, os.path.join(_save_class_path,
_new_xml_name))
# 继续循环处理
_deal_num += 1
# 返回结果
yield [_total, _total, True]
except:
print('labelimg_copy_flags_pics error: %s\r\n%s' %
(input_path, traceback.format_exc()))
yield [-1, -1, False]
