fix: refact logic

This commit is contained in:
2025-12-31 17:38:32 +08:00
parent 6ac50f7d2f
commit 35928c2484
17 changed files with 678 additions and 738 deletions

View File

@@ -1,122 +1,157 @@
"""DocLayout-YOLO wrapper for document layout detection."""
"""PP-DocLayoutV2 wrapper for document layout detection."""
import numpy as np
from app.schemas.image import LayoutInfo, LayoutRegion
from app.core.config import get_settings
from paddleocr import LayoutDetection
from typing import Optional
settings = get_settings()
class LayoutDetector:
"""Wrapper for DocLayout-YOLO model."""
"""Layout detector for PP-DocLayoutV2."""
# Class names from DocLayout-YOLO
CLASS_NAMES = {
0: "title",
1: "plain_text",
2: "abandon",
3: "figure",
4: "figure_caption",
5: "table",
6: "table_caption",
7: "table_footnote",
8: "isolate_formula",
9: "formula_caption",
_layout_detector: Optional[LayoutDetection] = None
# PP-DocLayoutV2 class ID to label mapping
CLS_ID_TO_LABEL: dict[int, str] = {
0: "abstract",
1: "algorithm",
2: "aside_text",
3: "chart",
4: "content",
5: "display_formula",
6: "doc_title",
7: "figure_title",
8: "footer",
9: "footer_image",
10: "footnote",
11: "formula_number",
12: "header",
13: "header_image",
14: "image",
15: "inline_formula",
16: "number",
17: "paragraph_title",
18: "reference",
19: "reference_content",
20: "seal",
21: "table",
22: "text",
23: "vertical_text",
24: "vision_footnote",
}
# Classes considered as plain text
PLAIN_TEXT_CLASSES = {"title", "plain_text", "figure_caption", "table_caption", "table_footnote"}
# Mapping from raw labels to normalized region types
LABEL_TO_TYPE: dict[str, str] = {
# Text types
"abstract": "text",
"algorithm": "text",
"aside_text": "text",
"content": "text",
"doc_title": "text",
"footer": "text",
"footnote": "text",
"header": "text",
"number": "text",
"paragraph_title": "text",
"reference": "text",
"reference_content": "text",
"text": "text",
"vertical_text": "text",
"vision_footnote": "text",
# Formula types
"display_formula": "formula",
"inline_formula": "formula",
"formula_number": "formula",
# Table types
"table": "table",
# Figure types
"chart": "figure",
"figure_title": "figure",
"footer_image": "figure",
"header_image": "figure",
"image": "figure",
"seal": "figure",
}
# Classes considered as formula
FORMULA_CLASSES = {"isolate_formula", "formula_caption"}
def __init__(self, model_path: str, confidence_threshold: float = 0.2):
"""Initialize the layout detector.
def __init__(self):
"""Initialize layout detector.
Args:
model_path: Path to the DocLayout-YOLO model weights.
confidence_threshold: Minimum confidence for detections.
"""
self.model_path = model_path
self.confidence_threshold = confidence_threshold
self.model = None
_ = self._get_layout_detector()
def load_model(self) -> None:
"""Load the DocLayout-YOLO model.
def _get_layout_detector(self):
"""Get or create LayoutDetection instance."""
if LayoutDetector._layout_detector is None:
LayoutDetector._layout_detector = LayoutDetection(model_name="PP-DocLayoutV2")
return LayoutDetector._layout_detector
Raises:
RuntimeError: If model cannot be loaded.
"""
try:
from doclayout_yolo import YOLOv10
self.model = YOLOv10(self.model_path)
except Exception as e:
raise RuntimeError(f"Failed to load DocLayout-YOLO model: {e}") from e
def detect(self, image: np.ndarray, image_size: int = 1024) -> LayoutInfo:
"""Detect document layout regions.
def detect(self, image: np.ndarray) -> LayoutInfo:
"""Detect layout of the image using PP-DocLayoutV2.
Args:
image: Input image as numpy array in BGR format.
image_size: Image size for prediction.
image: Input image as numpy array.
Returns:
LayoutInfo with detected regions.
Raises:
RuntimeError: If model not loaded.
LayoutInfo with detected regions and flags.
"""
if self.model is None:
raise RuntimeError("Model not loaded. Call load_model() first.")
# Run prediction
results = self.model.predict(
image,
imgsz=image_size,
conf=self.confidence_threshold,
device=settings.device,
)
layout_detector = self._get_layout_detector()
result = layout_detector.predict(image)
# Parse the result
regions: list[LayoutRegion] = []
has_plain_text = False
has_formula = False
mixed_recognition = False
if results and len(results) > 0:
result = results[0]
if result.boxes is not None:
for box in result.boxes:
cls_id = int(box.cls[0].item())
confidence = float(box.conf[0].item())
bbox = box.xyxy[0].tolist()
# Handle result format: [{'input_path': ..., 'page_index': None, 'boxes': [...]}]
if isinstance(result, list) and len(result) > 0:
first_result = result[0]
if isinstance(first_result, dict) and "boxes" in first_result:
boxes = first_result.get("boxes", [])
else:
boxes = []
else:
boxes = []
class_name = self.CLASS_NAMES.get(cls_id, f"unknown_{cls_id}")
for box in boxes:
cls_id = box.get("cls_id")
label = box.get("label") or self.CLS_ID_TO_LABEL.get(cls_id, "other")
score = box.get("score", 0.0)
coordinate = box.get("coordinate", [0, 0, 0, 0])
# Map to simplified type
if class_name in self.PLAIN_TEXT_CLASSES:
region_type = "text"
has_plain_text = True
elif class_name in self.FORMULA_CLASSES:
region_type = "formula"
has_formula = True
elif class_name in {"figure"}:
region_type = "figure"
elif class_name in {"table"}:
region_type = "table"
else:
region_type = class_name
# Normalize label to region type
region_type = self.LABEL_TO_TYPE.get(label, "text")
regions.append(
LayoutRegion(
type=region_type,
bbox=bbox,
confidence=confidence,
)
)
regions.append(LayoutRegion(
type=region_type,
bbox=coordinate,
confidence=score,
score=score,
))
return LayoutInfo(
regions=regions,
has_plain_text=has_plain_text,
has_formula=has_formula,
)
mixed_recognition = any(region.type == "text" and region.score > 0.85 for region in regions)
return LayoutInfo(regions=regions, MixedRecognition=mixed_recognition)
if __name__ == "__main__":
import cv2
from app.services.image_processor import ImageProcessor
layout_detector = LayoutDetector()
image_path = "test/timeout.png"
image = cv2.imread(image_path)
image_processor = ImageProcessor(padding_ratio=0.15)
image = image_processor.add_padding(image)
# Save the padded image for debugging
cv2.imwrite("debug_padded_image.png", image)
layout_info = layout_detector.detect(image)
print(layout_info)