YogeLiu/doc_processer
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: optimize the format convert

Browse Source
This commit is contained in:
liuyuanchuang
2026-02-04 12:00:06 +08:00
parent 10dbd59161
commit 526c1f3a0d
7 changed files with 571 additions and 187 deletions
Download Patch File Download Diff File Expand all files Collapse all files

119
app/services/ocr_service.py
View File

@@ -17,13 +17,31 @@ settings = get_settings()
_COMMANDS_NEED_SPACE = {
# operators / calculus
"cdot", "times", "div", "pm", "mp",
"int", "iint", "iiint", "oint", "sum", "prod", "lim",
"cdot",
"times",
"div",
"pm",
"mp",
"int",
"iint",
"iiint",
"oint",
"sum",
"prod",
"lim",
# common functions
"sin", "cos", "tan", "cot", "sec", "csc",
"log", "ln", "exp",
"sin",
"cos",
"tan",
"cot",
"sec",
"csc",
"log",
"ln",
"exp",
# misc
"partial", "nabla",
"partial",
"nabla",
}
_MATH_SEGMENT_PATTERN = re.compile(r"\$\$.*?\$\$|\$.*?\$", re.DOTALL)
@@ -58,7 +76,7 @@ def _split_glued_command_token(token: str) -> str:
if not best:
return token
suffix = body[len(best):]
suffix = body[len(best) :]
if not suffix:
return token
@@ -118,11 +136,11 @@ class OCRService(OCRServiceBase):
image_processor: Image processor instance.
"""
self.vl_server_url = vl_server_url or settings.paddleocr_vl_url
self.layout_detector = layout_detector
self.layout_detector = layout_detector
self.image_processor = image_processor
self.converter = converter
def _get_pipeline(self):
def _get_pipeline(self):
"""Get or create PaddleOCR-VL pipeline.
Returns:
@@ -159,12 +177,13 @@ class OCRService(OCRServiceBase):
markdown_content += res.markdown.get("markdown_texts", "")
markdown_content = _postprocess_markdown(markdown_content)
convert_result = self.converter.convert_to_formats(markdown_content)
convert_result = self.converter.convert_to_formats(markdown_content)
return {
"markdown": markdown_content,
"latex": convert_result.latex,
"mathml": convert_result.mathml,
"mml": convert_result.mml,
}
except Exception as e:
raise RuntimeError(f"Mixed recognition failed: {e}") from e
@@ -196,6 +215,7 @@ class OCRService(OCRServiceBase):
return {
"latex": convert_result.latex,
"mathml": convert_result.mathml,
"mml": convert_result.mml,
"markdown": markdown_content,
}
except Exception as e:
@@ -220,7 +240,7 @@ class OCRService(OCRServiceBase):
class MineruOCRService(OCRServiceBase):
"""Service for OCR using local file_parse API."""
def __init__(
self,
api_url: str = "http://127.0.0.1:8000/file_parse",
@@ -228,7 +248,7 @@ class MineruOCRService(OCRServiceBase):
converter: Optional[Converter] = None,
):
"""Initialize Local API service.
Args:
api_url: URL of the local file_parse API endpoint.
converter: Optional converter instance for format conversion.
@@ -236,13 +256,13 @@ class MineruOCRService(OCRServiceBase):
self.api_url = api_url
self.image_processor = image_processor
self.converter = converter
def recognize(self, image: np.ndarray) -> dict:
"""Recognize content using local file_parse API.
Args:
image: Input image as numpy array in BGR format.
Returns:
Dict with 'markdown', 'latex', 'mathml' keys.
"""
@@ -251,78 +271,71 @@ class MineruOCRService(OCRServiceBase):
image = self.image_processor.add_padding(image)
# Convert numpy array to image bytes
success, encoded_image = cv2.imencode('.png', image)
success, encoded_image = cv2.imencode(".png", image)
if not success:
raise RuntimeError("Failed to encode image")
image_bytes = BytesIO(encoded_image.tobytes())
# Prepare multipart form data
files = {
'files': ('image.png', image_bytes, 'image/png')
}
files = {"files": ("image.png", image_bytes, "image/png")}
data = {
'return_middle_json': 'false',
'return_model_output': 'false',
'return_md': 'true',
'return_images': 'false',
'end_page_id': '99999',
'start_page_id': '0',
'lang_list': 'en',
'server_url': 'string',
'return_content_list': 'false',
'backend': 'hybrid-auto-engine',
'table_enable': 'true',
'response_format_zip': 'false',
'formula_enable': 'true',
'parse_method': 'ocr'
"return_middle_json": "false",
"return_model_output": "false",
"return_md": "true",
"return_images": "false",
"end_page_id": "99999",
"start_page_id": "0",
"lang_list": "en",
"server_url": "string",
"return_content_list": "false",
"backend": "hybrid-auto-engine",
"table_enable": "true",
"response_format_zip": "false",
"formula_enable": "true",
"parse_method": "ocr",
}
# Make API request
response = requests.post(
self.api_url,
files=files,
data=data,
headers={'accept': 'application/json'},
timeout=30
)
response = requests.post(self.api_url, files=files, data=data, headers={"accept": "application/json"}, timeout=30)
response.raise_for_status()
result = response.json()
# Extract markdown content from response
markdown_content = ""
if 'results' in result and 'image' in result['results']:
markdown_content = result['results']['image'].get('md_content', '')
if "results" in result and "image" in result["results"]:
markdown_content = result["results"]["image"].get("md_content", "")
# markdown_content = _postprocess_markdown(markdown_content)
# Convert to other formats if converter is available
latex = ""
mathml = ""
mml = ""
if self.converter and markdown_content:
convert_result = self.converter.convert_to_formats(markdown_content)
latex = convert_result.latex
mathml = convert_result.mathml
mml = convert_result.mml
return {
"markdown": markdown_content,
"latex": latex,
"mathml": mathml,
"mml": mml,
}
except requests.RequestException as e:
raise RuntimeError(f"Local API request failed: {e}") from e
except Exception as e:
raise RuntimeError(f"Recognition failed: {e}") from e
if __name__ == "__main__":
mineru_service = MineruOCRService()
image = cv2.imread("test/complex_formula.png")
image_numpy = np.array(image)
ocr_result = mineru_service.recognize(image_numpy)
print(ocr_result)
print(ocr_result)