def to_mp3(cls,
input_file,
output_file,
output_ac,
output_ar,
input_ac=None,
input_ar=None,
input_format=None,
input_acodec=None):
"""
转 mp3 格式音频
:param str input_file: 待转换文件路径
:param str output_file:
:param int output_ac:
:param int output_ar:
:param int input_ac:
:param int input_ar:
:param str input_format:
:param str input_acodec:
:return:
"""
Assert.is_true(cls.is_mp3(output_file, is_strict=False),
'不是 mp3 拓展文件：{0}'.format(output_file))
cls._format_convert(input_file=input_file,
output_file=output_file,
input_format=input_format,
input_acodec=input_acodec,
input_ac=input_ac,
input_ar=input_ar,
output_ac=output_ac,
output_ar=output_ar)
def lexer(self, text, ne_list=[], url=None):
"""
词法分析
:param url:
:param text:
:param ne_list: 过滤命名实体类型
:return:
"""
url = url if url else getattr(self, 'lexer_url', None)
url = url.format(access_token=self.token)
res = requests.post(url=url,
data=json.dumps(obj={'text': text}),
headers=ContentType.JSON_UTF8.value)
res.encoding = encodes.Unicode.UTF_8.value
if res is None or res.status_code != codes.ok:
raise MyError(code=codes.failed, msg='百度词法分析请求失败.')
lexer_res, errors = LexerRes.LexerResSchema().load(
res.json()) # type: LexerRes
Assert.is_true(is_empty(errors), errors)
if is_not_empty(ne_list):
for lexer_item in lexer_res.items[:]:
if lexer_item.ne not in ne_list:
# 根据实体列表过滤
lexer_res.items.remove(lexer_item)
return lexer_res
def to_wav(cls,
input_file,
output_file,
output_ac,
output_ar,
input_ac=None,
input_ar=None,
input_format=None,
input_acodec=None):
"""
转 wav 格式音频
:param input_file:
:param output_file:
:param output_ac:
:param output_ar:
:param input_ac:
:param input_ar:
:param input_format:
:param input_acodec:
:return:
"""
Assert.is_true(cls.is_wav(output_file, is_strict=False),
'不是 wav 拓展文件：{0}'.format(output_file))
cls._format_convert(input_file=input_file,
output_file=output_file,
input_format=input_format,
input_acodec=input_acodec,
input_ac=input_ac,
input_ar=input_ar,
output_ac=output_ac,
output_ar=output_ar)
def get_asr_progress(self):
"""
查询音频转写处理进度
:return:
"""
url = getattr(self, 'asr_progress_url', None)
data = dict(app_id=self.app_id,
signa=self.signa,
ts=self.ts,
task_id=self.task_id)
res = requests.post(url=url, data=data)
res.encoding = encodes.Unicode.UTF_8.value
if res is None or res.status_code != codes.ok:
return AsrInfo(
ok=codes.failed,
err_no=res.status_code if res is not None else codes.bad,
failed='查询音频转写处理进度请求失败')
progress_info, errors = AsrInfo.AsrInfoSchema().load(res.json())
Assert.is_true(is_empty(errors), errors)
if progress_info.ok == codes.success:
progress_details, errors = AsrProgress.AsrProgressSchema().load(
json.loads(progress_info.data))
Assert.is_true(is_empty(errors), errors)
progress_info.data = progress_details
return progress_info
def __init__(self, key):
"""
:param key: 秘钥
"""
Assert.is_true(is_not_empty(key), 'AES 秘钥不能为空')
self.key = key
def probe(cls, file_path, is_ffprobe=True):
"""
解析音频文件参数
:param is_ffprobe: 是否使用 ffprobe 解析文件格式
:param file_path:
:return:
:rtype: Audio
"""
Assert.is_true(os.path.isfile(file_path),
'文件不存在：{0}'.format(file_path))
_, file_name, format_name = FileUtil.get_path_name_ext(file_path)
probe_json = ffmpeg.probe(file_path)
audio_json = probe_json.get('streams')[0]
audio_json.update(probe_json.get('format'))
audio_json.update({
'file_name':
'{name}.{ext}'.format(name=file_name, ext=format_name)
})
audio, errors = AudioSchema().load(audio_json)
Assert.is_true(is_empty(errors), errors)
if not is_ffprobe:
audio.format_name = format_name
return audio
def get_file_list(self, path):
"""
获取文件列表
:param path:文件目录所在文件夹
:return:文件对象组成的元组,注意返回结果里面包含所有文件甚至是 . 和 .. 也会包含进去
"""
Assert.is_true(assert_condition=self.is_success(), assert_msg='SMB连接失败...')
return self.listPath(service_name=self.service_name, path=path)
def del_file(file_path):
"""
删除文件
:param file_path:
:return:
"""
Assert.is_true(os.path.isfile(file_path),
'文件不存在：{0}'.format(file_path))
os.remove(file_path)
def get_file_create_time(file_path):
"""
获取文件创建时间
:param file_path:
:return:
"""
Assert.is_true(os.path.isfile(file_path),
'文件不存在, path: {0}'.format(file_path))
return os.path.getctime(file_path)
def get_file_size(file_path):
"""
获取文件大小 保留两位小数，单位MB
:param file_path:
:return:
"""
Assert.is_true(os.path.isfile(file_path),
'文件不存在, path: {0}'.format(file_path))
return round(os.path.getsize(file_path) / float(1024 * 1024), 2)
def get_img_pixel(path):
"""
获取图片像素大小
:param path:
:return:
"""
Assert.is_true(os.path.isfile(path), '图片不存在：{0}'.format(path))
with Image.open(path) as i:
return i.size
def get_file_access_time(file_path):
"""
获取文件访问时间
:param file_path:
:return:
"""
Assert.is_true(os.path.isfile(file_path),
'文件不存在, path: {0}'.format(file_path))
return os.path.getatime(file_path)
def upload_file(self, local_file, remote_file):
"""
上传文件到服务器
:param local_file:本地文件路径
:param remote_file:远程文件路径
:return:
"""
Assert.is_true(assert_condition=self.is_success(), assert_msg='SMB连接失败...')
with open(file=local_file, mode='rb') as f:
self.storeFile(service_name=self.service_name, path=remote_file, file_obj=f)
def down_file(self, local_file, remote_file):
"""
下载文件到本地
:param local_file:本地文件路径
:param remote_file:远程文件路径
:return:
"""
Assert.is_true(assert_condition=self.is_success(), assert_msg='SMB连接失败...')
with open(file=local_file, mode='wb') as f:
self.retrieveFile(service_name=self.service_name, path=remote_file, file_obj=f)
def creat_dirs(dir_path):
"""
创建多级目录
:param dir_path:
:return:
"""
Assert.is_true(dir_path, '目录不能为空')
dir_path = dir_path.strip()
if not os.path.exists(dir_path):
os.makedirs(dir_path)
Assert.is_true(os.path.isdir(dir_path), '创建目录失败：{0}'.format(dir_path))
return dir_path
def loss_less(in_path, out_path, format=FileFormat.JPEG.value):
"""
图片近无损压缩
:param in_path:
:param out_path: 输出目录
:param format:
:return:
"""
Assert.is_true(os.path.isfile(in_path), '图片不存在：{0}'.format(in_path))
with Image.open(in_path) as img:
img = img.convert(ImgChannel.RGB.value)
exif_bytes = piexif.dump({})
img.save(out_path, format, exif=exif_bytes)
def copy_file(cls, old_file, new_file):
"""
复制文件到指定目录
:param old_file:源文件路径
:param new_file:新文件目录
:return:
"""
Assert.is_true(os.path.isfile(old_file), '文件不存在：{0}'.format(old_file))
file_name = os.path.basename(old_file)
FileUtil.creat_dirs(new_file)
new_file_path = os.path.join(new_file, file_name)
shutil.copyfile(src=old_file, dst=new_file_path)
return cls.normcase(new_file_path)
def get_path_name_ext(file_path, is_strict=True):
"""
解析文件路径，文件名，文件扩展
:param bool is_strict: 是否严格检测文件是否存在
:param file_path:
:return:
"""
if is_strict:
Assert.is_true(os.path.isfile(file_path),
'文件不存在, path: {0}'.format(file_path))
(path, file_name) = os.path.split(file_path)
(name, ext) = os.path.splitext(file_name)
return path, name, ext.strip(os.curdir)
def get_files_by_suffix2(dir, suffixs):
"""
获取目录下所有符合后缀条件的文件路径列表
:param dir:目录
:param suffixs:后缀
:return:
"""
Assert.is_true(assert_condition=os.path.isdir(dir),
assert_msg='路径无效：{0}'.format(dir))
file_list = []
for suffix in suffixs:
file_glob = os.path.join(dir, '*.' + suffix)
file_list.extend(glob.glob(pathname=file_glob))
return file_list
def _format_convert(cls,
input_file,
output_file,
output_ac,
output_ar,
input_ac=None,
input_ar=None,
input_format=None,
input_acodec=None,
output_format=None,
output_acodec=None):
"""
音频文件格式转换
:param input_file:
:param output_file:
:param output_ac:
:param output_ar:
:param input_ac:
:param input_ar:
:param input_format:
:param input_acodec:
:param output_format:
:param output_acodec:
:return:
"""
output_dir, _, _ = FileUtil.get_path_name_ext(output_file,
is_strict=False)
FileUtil.creat_dirs(output_dir)
if cls.is_pcm(input_file):
# 格式为 pcm 格式，需要音频文件参数
Assert.is_true((is_not_empty(input_ac) and is_not_empty(input_ar)
and is_not_empty(input_format)
and is_not_empty(input_acodec)), 'pcm 缺失文件参数')
_, error = (ffmpeg.input(
input_file,
**cls._file_args(
input_format, input_acodec, input_ac, input_ar)).output(
output_file,
**cls._file_args(output_format, output_acodec, output_ac,
output_ar)).run(capture_stdout=False,
overwrite_output=True))
def img_compress(path, threshold=4):
"""
将图片压缩到指定阀值大小的 base64
:param path:文件路径
:param threshold:阀值大小(单位：M)
:return:
"""
from utils.encodes import pil_to_base64
# 阈值换算成比特
_threshold = threshold * 1024 * 1024
Assert.is_true(os.path.isfile(path), '图片不存在：{0}'.format(path))
w, h = ImgUtil.get_img_pixel(path)
with Image.open(path) as im:
if w * h > _threshold:
new_width = 1024
new_height = int(new_width * h * 1.0 / w)
resized_im = im.resize((new_width, new_height))
return pil_to_base64(resized_im)
return pil_to_base64(im)
def _asr_prepare(self, **kwargs):
"""
音频转写预处理
:return:
"""
url = getattr(self, 'asr_prepare_url', None)
kwargs.update(dict(app_id=self.app_id, signa=self.signa, ts=self.ts))
res = requests.post(url=url, data=kwargs)
res.encoding = encodes.Unicode.UTF_8.value
if res is None or res.status_code != codes.ok:
return AsrInfo(
ok=codes.failed,
err_no=res.status_code if res is not None else codes.bad,
failed='音频转写预处理请求失败')
prepare_info, errors = AsrInfo.AsrInfoSchema().load(res.json())
Assert.is_true(is_empty(errors), errors)
return prepare_info
def get_asr_result(self):
"""
查询音频转写结果
:return:
"""
url = getattr(self, 'asr_result_url', None)
data = dict(app_id=self.app_id,
signa=self.signa,
ts=self.ts,
task_id=self.task_id)
res = requests.post(url=url, data=data)
res.encoding = encodes.Unicode.UTF_8.value
if res is None or res.status_code != codes.ok:
return AsrInfo(
ok=codes.failed,
err_no=res.status_code if res is not None else codes.bad,
failed='查询音频转写结果请求失败')
result_info, errors = AsrInfo.AsrInfoSchema().load(res.json())
Assert.is_true(is_empty(errors), errors)
return result_info
def _asr_upload(self, upload_file_path):
"""
音频转写文件分片上传
:param upload_file_path: 待上传音频文件路径
:return:
"""
url = getattr(self, 'asr_upload_url', None)
sig = SliceIdGenerator()
with open(upload_file_path, 'rb') as f:
while True:
content = f.read(self.__FILE_PIECE_SIZE)
if not content or len(content) == 0:
break
data = dict(app_id=self.app_id,
signa=self.signa,
ts=self.ts,
task_id=self.task_id,
slice_id=sig.get_next_slice_id())
res = requests.post(url=url,
data=data,
files={'content': content})
res.encoding = encodes.Unicode.UTF_8.value
if res is None or res.status_code != codes.ok:
return AsrInfo(ok=codes.failed,
err_no=res.status_code
if res is not None else codes.bad,
failed='音频转写文件分片上传请求失败')
upload_info, errors = AsrInfo.AsrInfoSchema().load(res.json())
Assert.is_true(is_empty(errors), errors)
# 上传分片失败
Assert.is_true(upload_info.ok == codes.success,
upload_info.err_no,
'分片上传失败：{0}'.format(upload_info.failed))
def asr(self,
file_path,
has_participle=False,
has_seperate=True,
speaker_number=2):
"""
音频转写
:param str file_path: 音频文件路径
:param bool has_participle: 转写结果是否包含分词信息
:param bool has_seperate: 转写结果中是否包含发音人分离信息
:param int speaker_number: 发音人个数，可选值：0-10，0表示盲分
:return:
"""
from utils.file import Audio as BasicAudio
has_participle = 'true' if has_participle else 'false'
has_seperate = 'true' if has_seperate else 'false'
audio_file = BasicAudio.probe(file_path)
# 音频转写预处理
prepare_info = self._asr_prepare(
file_len=audio_file.size,
file_name=audio_file.file_name,
slice_num=math.ceil(audio_file.size / self.__FILE_PIECE_SIZE),
has_participle=has_participle,
has_seperate=has_seperate,
speaker_number=speaker_number)
Assert.is_true(prepare_info.ok == codes.success, prepare_info.failed,
prepare_info.err_no)
# 预处理成功，更新task_id
self.task_id = prepare_info.data
# 文件发片上传
self._asr_upload(file_path)
# 文件合并
merge_info = self._asr_merge()
Assert.is_true(merge_info.ok == codes.success, merge_info.failed,
merge_info.err_no)
def to_pcm(cls,
input_file,
output_file,
output_ac,
output_ar,
input_ac=None,
input_ar=None,
input_format=None,
input_acodec=None,
output_format=AudioFormat.S16LE.value,
output_acodec=AudioCodecs.PCM_S16LE.value):
"""
转 pcm 格式音频
:param str input_file: 待转换文件路径
:param int input_ac: 输入声道数（pcm 格式需要）
:param int input_ar: 输入采样率 （pcm 格式需要）
:param str input_format: 输入格式 （pcm 格式需要）
:param str input_acodec: 输入编码器 （pcm 格式需要）
:param str output_file: 转换输出文件路径
:param int output_ac: 输出声道数
:param int output_ar: 输出采样率
:param str output_format: 输出格式
:param str output_acodec: 输出编码器
:return:
"""
Assert.is_true(cls.is_pcm(output_file, is_strict=False),
'不是 pcm 拓展文件：{0}'.format(output_file))
cls._format_convert(input_file=input_file,
output_file=output_file,
input_format=input_format,
input_acodec=input_acodec,
input_ac=input_ac,
input_ar=input_ar,
output_format=output_format,
output_acodec=output_acodec,
output_ac=output_ac,
output_ar=output_ar)
def get_md5_path(file_path,
block_size=64 * 1024,
hexadecimal=True,
b64=False):
"""
获取文件 md5
:param file_path: 文件路径
:param block_size: 批次读取size大小
:param hexadecimal: 是否使用十六进制，否则使用二进制
:param b64: 是否 base64 编码
:return:
"""
from utils.encodes import hash_code
Assert.is_true(os.path.isfile(file_path),
'文件不存在, path: {0}'.format(file_path))
with open(file_path, 'rb') as f:
md5 = hashlib.md5()
while True:
data = f.read(block_size)
if not data:
break
md5.update(data)
return hash_code(md5, hexadecimal, b64)
def pdf_to_pic(path,
pic_dir,
format=FileFormat.JPG.value,
loss=True,
gamma=True,
zoom=210,
min_size=1.50,
max_size=15.0):
"""
从pdf中提取图片
:param path: pdf的路径
:param pic_dir: 图片保存的路径
:param format: 图片格式
:param bool loss: 是否压缩
:param bool gamma: 是否 gamma 矫正
:param int zoom: 保存图片分辨率
:param min_size: 文件大小最小阈值
:param max_size: 文件大小最大阈值
:return: {page_num, success_num, fail_num, msg}
"""
Assert.is_true(os.path.isfile(path), '文件不存在, path: {0}'.format(path))
page_num, success_num, fail_num = 0, 0, 0
detail_info = {'images': []}
pdf = None
# 初始化图片压缩
pngquant = PngQuant(min_quality=80,
max_quality=100,
tmp_file=os.path.join(pic_dir, 'quant.tmp.png'))
try:
FileUtil.creat_dirs(pic_dir)
pdf = fitz.Document(path)
page_num = pdf.pageCount
for pg in range(page_num):
pg_zoom = zoom
pm_dict = {'page_code': pg + 1}
try:
page = pdf[pg] # type: Page
trans = fitz.Matrix(pg_zoom / 100.0,
pg_zoom / 100.0).preRotate(0)
pm = page.getPixmap(matrix=trans, alpha=False) # 获得每一页的流对象
page_path = FileUtil.path_join(pic_dir, '{0}.{1}'.format(
(pg + 1), format)) # 图片路径
pm.writeImage(page_path) # 保存图片
file_size = FileUtil.get_file_size(page_path)
if file_size <= min_size:
# 低于最小阀值
pg_zoom = zoom * 2
trans = fitz.Matrix(pg_zoom / 100.0,
pg_zoom / 100.0).preRotate(0)
pm = page.getPixmap(matrix=trans, alpha=False)
pm.writeImage(page_path)
elif file_size >= max_size:
# 大于最大阀值
pg_zoom = zoom * 2 / 3
trans = fitz.Matrix(pg_zoom / 100.0,
pg_zoom / 100.0).preRotate(0)
pm = page.getPixmap(matrix=trans, alpha=False)
pm.writeImage(page_path)
if gamma:
# gamma 矫正
Enhancer().enhance(page_path, False, False, False,
False, True)
if loss:
# 对图片进行近无损压缩
pngquant.quant_image(page_path)
pm_dict['img_path'] = page_path
success_num = success_num + 1
except Exception as e:
page_path = pm_dict.get('img_path', '')
if is_not_empty(page_path):
pm_dict.pop('img_path')
if os.path.isfile(page_path):
# 处理失败，删除失败文件
FileUtil.del_file(page_path)
pm_dict['error_msg'] = repr(e) if repr(e) else 'pdf转图片失败'
fail_num = fail_num + 1
finally:
detail_info.get('images').append(pm_dict)
except Exception as e1:
detail_info['error_msg'] = repr(e1) if repr(e1) else '处理pdf失败'
finally:
if is_not_empty(pdf):
pdf.close()
if success_num + fail_num != page_num:
fail_num = page_num - success_num
return page_num, success_num, fail_num, detail_info
