BabelDOC 0.3.71__tar.gz → 0.4.0__tar.gz

{babeldoc-0.3.71 → babeldoc-0.4.0}/.gitignore

@@ -33,4 +33,5 @@ examples/
 .cursor/rules/_*.mdc
 /.cursor
 /xnotes
-/docs/workflow-rules.md
+/docs/workflow-rules.md
+babeldoc/format/txt

{babeldoc-0.3.71 → babeldoc-0.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: BabelDOC
-Version: 0.3.71
+Version: 0.4.0
 Summary: Yet Another Document Translator
 Project-URL: Homepage, https://github.com/funstory-ai/BabelDOC
 Project-URL: Issues, https://github.com/funstory-ai/BabelDOC/issues
@@ -17,6 +17,7 @@ Requires-Dist: charset-normalizer>=2.0.0
 Requires-Dist: configargparse>=1.7
 Requires-Dist: cryptography>=36.0.0
 Requires-Dist: freetype-py>=2.5.1
+Requires-Dist: hdbscan>=0.8.40
 Requires-Dist: httpx[socks]>=0.27.0
 Requires-Dist: huggingface-hub>=0.27.0
 Requires-Dist: hyperscan>=0.7.13
@@ -35,7 +36,9 @@ Requires-Dist: python-levenshtein>=0.27.1
 Requires-Dist: pyzstd>=0.17.0
 Requires-Dist: rapidocr-onnxruntime>=1.4.4
 Requires-Dist: rich>=13.9.4
+Requires-Dist: rtree>=1.4.0
 Requires-Dist: scikit-image>=0.25.2
+Requires-Dist: scikit-learn<1.8.0
 Requires-Dist: tenacity>=9.0.0
 Requires-Dist: tiktoken>=0.9.0
 Requires-Dist: toml>=0.10.2
@@ -371,7 +374,7 @@ You can refer to the example in [main.py](https://github.com/funstory-ai/yadt/bl
 
 Please note:
 
-1. Make sure call `babeldoc.high_level.init()` before using the API
+1. Make sure call `babeldoc.format.pdf.high_level.init()` before using the API
 
 2. The current `TranslationConfig` does not fully validate input parameters, so you need to ensure the validity of input parameters
 
@@ -449,6 +452,14 @@ And meet the following requirements:
 - layout error less than 1%
 - content loss less than 1%
 
+## Version Number Explanation
+
+This project uses a combination of [Semantic Versioning](https://semver.org/) and [Pride Versioning](https://pridever.org/). The version number format is: "0.MAJOR.MINOR".
+
+- MAJOR: Incremented by 1 when API incompatible changes are made or when proud improvements are implemented.
+
+- MINOR: Incremented by 1 when any API compatible changes are made.
+
 ## Known Issues
 
 1. Parsing errors in the author and reference sections; they get merged into one paragraph after translation.

{babeldoc-0.3.71 → babeldoc-0.4.0}/README.md

@@ -320,7 +320,7 @@ You can refer to the example in [main.py](https://github.com/funstory-ai/yadt/bl
 
 Please note:
 
-1. Make sure call `babeldoc.high_level.init()` before using the API
+1. Make sure call `babeldoc.format.pdf.high_level.init()` before using the API
 
 2. The current `TranslationConfig` does not fully validate input parameters, so you need to ensure the validity of input parameters
 
@@ -398,6 +398,14 @@ And meet the following requirements:
 - layout error less than 1%
 - content loss less than 1%
 
+## Version Number Explanation
+
+This project uses a combination of [Semantic Versioning](https://semver.org/) and [Pride Versioning](https://pridever.org/). The version number format is: "0.MAJOR.MINOR".
+
+- MAJOR: Incremented by 1 when API incompatible changes are made or when proud improvements are implemented.
+
+- MINOR: Incremented by 1 when any API compatible changes are made.
+
 ## Known Issues
 
 1. Parsing errors in the author and reference sections; they get merged into one paragraph after translation.

babeldoc-0.4.0/babeldoc/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.4.0"

{babeldoc-0.3.71 → babeldoc-0.4.0}/babeldoc/const.py

@@ -3,7 +3,7 @@ import shutil
 import subprocess
 from pathlib import Path
 
-__version__ = "0.3.71"
+__version__ = "0.4.0"
 
 CACHE_FOLDER = Path.home() / ".cache" / "babeldoc"
 

{babeldoc-0.3.71 → babeldoc-0.4.0}/babeldoc/docvision/base_doclayout.py

@@ -4,7 +4,7 @@ from collections.abc import Generator
 
 import pymupdf
 
-from babeldoc.document_il.il_version_1 import Page
+from babeldoc.format.pdf.document_il.il_version_1 import Page
 
 logger = logging.getLogger(__name__)
 

{babeldoc-0.3.71 → babeldoc-0.4.0}/babeldoc/docvision/doclayout.py

@@ -8,9 +8,9 @@ from collections.abc import Generator
 import cv2
 import numpy as np
 
-from babeldoc.document_il.utils.mupdf_helper import get_no_rotation_img
 from babeldoc.docvision.base_doclayout import DocLayoutModel
 from babeldoc.docvision.base_doclayout import YoloResult
+from babeldoc.format.pdf.document_il.utils.mupdf_helper import get_no_rotation_img
 
 try:
     import onnx
@@ -24,7 +24,7 @@ except ImportError as e:
     raise
 import pymupdf
 
-import babeldoc.document_il.il_version_1
+import babeldoc.format.pdf.document_il.il_version_1
 from babeldoc.assets.assets import get_doclayout_onnx_model_path
 
 # from huggingface_hub import hf_hub_download
@@ -205,12 +205,12 @@ class OnnxModel(DocLayoutModel):
 
     def handle_document(
         self,
-        pages: list[babeldoc.document_il.il_version_1.Page],
+        pages: list[babeldoc.format.pdf.document_il.il_version_1.Page],
         mupdf_doc: pymupdf.Document,
         translate_config,
         save_debug_image,
     ) -> Generator[
-        tuple[babeldoc.document_il.il_version_1.Page, YoloResult], None, None
+        tuple[babeldoc.format.pdf.document_il.il_version_1.Page, YoloResult], None, None
     ]:
         for page in pages:
             translate_config.raise_if_cancelled()

{babeldoc-0.3.71 → babeldoc-0.4.0}/babeldoc/docvision/rpc_doclayout.py

@@ -14,10 +14,10 @@ from tenacity import stop_after_attempt
 from tenacity import wait_exponential
 
 import babeldoc
-from babeldoc.document_il.utils.mupdf_helper import get_no_rotation_img
 from babeldoc.docvision.base_doclayout import DocLayoutModel
 from babeldoc.docvision.base_doclayout import YoloBox
 from babeldoc.docvision.base_doclayout import YoloResult
+from babeldoc.format.pdf.document_il.utils.mupdf_helper import get_no_rotation_img
 
 logger = logging.getLogger(__name__)
 
@@ -270,7 +270,7 @@ class RpcDocLayoutModel(DocLayoutModel):
 
     def handle_document(
         self,
-        pages: list[babeldoc.document_il.il_version_1.Page],
+        pages: list[babeldoc.format.pdf.document_il.il_version_1.Page],
         mupdf_doc: pymupdf.Document,
         translate_config,
         save_debug_image,

babeldoc-0.4.0/babeldoc/docvision/rpc_doclayout2.py

@@ -0,0 +1,320 @@
+import logging
+import threading
+from concurrent.futures import ThreadPoolExecutor
+from pathlib import Path
+
+import cv2
+import httpx
+import msgpack
+import numpy as np
+import pymupdf
+from tenacity import retry
+from tenacity import retry_if_exception_type
+from tenacity import stop_after_attempt
+from tenacity import wait_exponential
+
+import babeldoc
+from babeldoc.docvision.base_doclayout import DocLayoutModel
+from babeldoc.docvision.base_doclayout import YoloBox
+from babeldoc.docvision.base_doclayout import YoloResult
+from babeldoc.format.pdf.document_il.utils.mupdf_helper import get_no_rotation_img
+
+logger = logging.getLogger(__name__)
+
+
+def encode_image(image) -> bytes:
+    """Read and encode image to bytes
+
+    Args:
+        image: Can be either a file path (str) or numpy array
+    """
+    if isinstance(image, str):
+        if not Path(image).exists():
+            raise FileNotFoundError(f"Image file not found: {image}")
+        img = cv2.imread(image)
+        if img is None:
+            raise ValueError(f"Failed to read image: {image}")
+    else:
+        img = image
+
+    # logger.debug(f"Image shape: {img.shape}")
+    encoded = cv2.imencode(".jpg", img)[1].tobytes()
+    # logger.debug(f"Encoded image size: {len(encoded)} bytes")
+    return encoded
+
+
+@retry(
+    stop=stop_after_attempt(3),  # 最多重试 3 次
+    wait=wait_exponential(
+        multiplier=1, min=1, max=10
+    ),  # 指数退避策略，初始 1 秒，最大 10 秒
+    retry=retry_if_exception_type((httpx.HTTPError, Exception)),  # 针对哪些异常重试
+    before_sleep=lambda retry_state: logger.warning(
+        f"Request failed, retrying in {retry_state.next_action.sleep} seconds... "
+        f"(Attempt {retry_state.attempt_number}/3)"
+    ),
+)
+def predict_layout(
+    image,
+    host: str = "http://localhost:8000",
+    imgsz: int = 1024,
+):
+    """
+    Predict document layout using the MOSEC service
+
+    Args:
+        image: Can be either a file path (str) or numpy array
+        host: Service host URL
+        imgsz: Image size for model input
+
+    Returns:
+        List of predictions containing bounding boxes and classes
+    """
+    # Prepare request data
+    if not isinstance(image, list):
+        image = [image]
+    image_data = [encode_image(image) for image in image]
+    data = {
+        "image": image_data,
+        "imgsz": imgsz,
+    }
+
+    # Pack data using msgpack
+    packed_data = msgpack.packb(data, use_bin_type=True)
+    # logger.debug(f"Packed data size: {len(packed_data)} bytes")
+
+    # Send request
+    # logger.debug(f"Sending request to {host}/inference")
+    response = httpx.post(
+        f"{host}/inference",
+        data=packed_data,
+        headers={
+            "Content-Type": "application/msgpack",
+            "Accept": "application/msgpack",
+        },
+        timeout=300,
+        follow_redirects=True,
+    )
+
+    # logger.debug(f"Response status: {response.status_code}")
+    # logger.debug(f"Response headers: {response.headers}")
+
+    if response.status_code == 200:
+        try:
+            result = msgpack.unpackb(response.content, raw=False)
+            if isinstance(result, dict):
+                names = {}
+                for box in result["boxes"]:
+                    box["xyxy"] = box["coordinate"]
+                    box["conf"] = box["score"]
+                    box["cls"] = box["cls_id"]
+                    names[box["cls_id"]] = box["label"]
+                if "names" not in result:
+                    result["names"] = names
+                result = [result]
+            return result
+        except Exception as e:
+            logger.exception(f"Failed to unpack response: {e!s}")
+            raise
+    else:
+        logger.error(f"Request failed with status {response.status_code}")
+        logger.error(f"Response content: {response.content}")
+        raise Exception(
+            f"Request failed with status {response.status_code}: {response.text}",
+        )
+
+
+class ResultContainer:
+    def __init__(self):
+        self.result = YoloResult(boxes_data=np.array([]), names=[])
+
+
+class RpcDocLayoutModel(DocLayoutModel):
+    """DocLayoutModel implementation that uses RPC service."""
+
+    def __init__(self, host: str = "http://localhost:8000"):
+        """Initialize RPC model with host address."""
+        self.host = host
+        self._stride = 32  # Default stride value
+        self._names = ["text", "title", "list", "table", "figure"]
+        self.lock = threading.Lock()
+
+    @property
+    def stride(self) -> int:
+        """Stride of the model input."""
+        return self._stride
+
+    def resize_and_pad_image(self, image, new_shape):
+        """
+        Resize and pad the image to the specified size,
+        ensuring dimensions are multiples of stride.
+
+        Parameters:
+        - image: Input image
+        - new_shape: Target size (integer or (height, width) tuple)
+        - stride: Padding alignment stride, default 32
+
+        Returns:
+        - Processed image
+        """
+        if isinstance(new_shape, int):
+            new_shape = (new_shape, new_shape)
+
+        h, w = image.shape[:2]
+        new_h, new_w = new_shape
+
+        # Calculate scaling ratio
+        r = min(new_h / h, new_w / w)
+        resized_h, resized_w = int(round(h * r)), int(round(w * r))
+
+        # Resize image
+        image = cv2.resize(
+            image, (resized_w, resized_h), interpolation=cv2.INTER_LINEAR
+        )
+
+        # Calculate padding size
+        pad_h = new_h - resized_h
+        pad_w = new_w - resized_w
+        top, bottom = pad_h // 2, pad_h - pad_h // 2
+        left, right = pad_w // 2, pad_w - pad_w // 2
+
+        # Add padding
+        image = cv2.copyMakeBorder(
+            image, top, bottom, left, right, cv2.BORDER_CONSTANT, value=(114, 114, 114)
+        )
+
+        return image
+
+    def scale_boxes(self, img1_shape, boxes, img0_shape):
+        """
+        Rescales bounding boxes (in the format of xyxy by default) from the shape of the image they were originally
+        specified in (img1_shape) to the shape of a different image (img0_shape).
+
+        Args:
+            img1_shape (tuple): The shape of the image that the bounding boxes are for,
+                in the format of (height, width).
+            boxes (torch.Tensor): the bounding boxes of the objects in the image, in the format of (x1, y1, x2, y2)
+            img0_shape (tuple): the shape of the target image, in the format of (height, width).
+
+        Returns:
+            boxes (torch.Tensor): The scaled bounding boxes, in the format of (x1, y1, x2, y2)
+        """
+
+        # Calculate scaling ratio
+        gain = min(img1_shape[0] / img0_shape[0], img1_shape[1] / img0_shape[1])
+
+        # Calculate padding size
+        pad_x = round((img1_shape[1] - img0_shape[1] * gain) / 2 - 0.1)
+        pad_y = round((img1_shape[0] - img0_shape[0] * gain) / 2 - 0.1)
+
+        # Remove padding and scale boxes
+        boxes = (boxes - [pad_x, pad_y, pad_x, pad_y]) / gain
+        return boxes
+
+    def predict_image(
+        self,
+        image,
+        host: str = None,
+        result_container: ResultContainer | None = None,
+        imgsz: int = 1024,
+    ) -> ResultContainer:
+        """Predict the layout of document pages using RPC service."""
+        if result_container is None:
+            result_container = ResultContainer()
+        target_imgsz = (800, 800)
+        orig_h, orig_w = image.shape[:2]
+        target_imgsz = (orig_h, orig_w)
+        if image.shape[0] != target_imgsz[0] or image.shape[1] != target_imgsz[1]:
+            image = self.resize_and_pad_image(image, new_shape=target_imgsz)
+        preds = predict_layout([image], host=self.host, imgsz=800)
+
+        if len(preds) > 0:
+            for pred in preds:
+                boxes = [
+                    YoloBox(
+                        None,
+                        self.scale_boxes(
+                            target_imgsz, np.array(x["xyxy"]), (orig_h, orig_w)
+                        ),
+                        np.array(x["conf"]),
+                        x["cls"],
+                    )
+                    for x in pred["boxes"]
+                ]
+                result_container.result = YoloResult(
+                    boxes=boxes,
+                    names={int(k): v for k, v in pred["names"].items()},
+                )
+        return result_container.result
+
+    def predict(self, image, imgsz=1024, **kwargs) -> list[YoloResult]:
+        """Predict the layout of document pages using RPC service."""
+        # Handle single image input
+        if isinstance(image, np.ndarray) and len(image.shape) == 3:
+            image = [image]
+
+        result_containers = [ResultContainer() for _ in image]
+        predict_thread = ThreadPoolExecutor(max_workers=len(image))
+        for img, result_container in zip(image, result_containers, strict=True):
+            predict_thread.submit(
+                self.predict_image, img, self.host, result_container, 800
+            )
+        predict_thread.shutdown(wait=True)
+        result = [result_container.result for result_container in result_containers]
+        return result
+
+    def predict_page(
+        self, page, mupdf_doc: pymupdf.Document, translate_config, save_debug_image
+    ):
+        translate_config.raise_if_cancelled()
+        with self.lock:
+            # pix = mupdf_doc[page.page_number].get_pixmap(dpi=72)
+            pix = get_no_rotation_img(mupdf_doc[page.page_number])
+        image = np.frombuffer(pix.samples, np.uint8).reshape(
+            pix.height,
+            pix.width,
+            3,
+        )[:, :, ::-1]
+        predict_result = self.predict_image(image, self.host, None, 800)
+        save_debug_image(image, predict_result, page.page_number + 1)
+        return page, predict_result
+
+    def handle_document(
+        self,
+        pages: list[babeldoc.format.pdf.document_il.il_version_1.Page],
+        mupdf_doc: pymupdf.Document,
+        translate_config,
+        save_debug_image,
+    ):
+        with ThreadPoolExecutor(max_workers=64) as executor:
+            yield from executor.map(
+                self.predict_page,
+                pages,
+                (mupdf_doc for _ in range(len(pages))),
+                (translate_config for _ in range(len(pages))),
+                (save_debug_image for _ in range(len(pages))),
+            )
+
+    @staticmethod
+    def from_host(host: str) -> "RpcDocLayoutModel":
+        """Create RpcDocLayoutModel from host address."""
+        return RpcDocLayoutModel(host=host)
+
+
+if __name__ == "__main__":
+    logging.basicConfig(level=logging.DEBUG)
+    # Test the service
+    try:
+        # Use a default test image if example/1.png doesn't exist
+        image_path = "example/1.png"
+        if not Path(image_path).exists():
+            print(f"Warning: {image_path} not found.")
+            print("Please provide the path to a test image:")
+            image_path = input("> ")
+
+        logger.info(f"Processing image: {image_path}")
+        result = predict_layout(image_path)
+        print("Prediction results:")
+        print(result)
+    except Exception as e:
+        print(f"Error: {e!s}")

{babeldoc-0.3.71 → babeldoc-0.4.0}/babeldoc/docvision/table_detection/rapidocr.py

@@ -6,9 +6,9 @@ from collections.abc import Generator
 import cv2
 import numpy as np
 from babeldoc.assets.assets import get_table_detection_rapidocr_model_path
-from babeldoc.document_il.utils.mupdf_helper import get_no_rotation_img
 from babeldoc.docvision.base_doclayout import YoloBox
 from babeldoc.docvision.base_doclayout import YoloResult
+from babeldoc.format.pdf.document_il.utils.mupdf_helper import get_no_rotation_img
 from rapidocr_onnxruntime import RapidOCR
 
 try:
@@ -20,7 +20,7 @@ except ImportError as e:
             "Download it at https://aka.ms/vs/17/release/vc_redist.x64.exe"
         ) from e
     raise
-import babeldoc.document_il.il_version_1
+import babeldoc.format.pdf.document_il.il_version_1
 import pymupdf
 
 logger = logging.getLogger(__name__)
@@ -96,7 +96,7 @@ class RapidOCRModel:
         self.model = RapidOCR(
             det_model_path=get_table_detection_rapidocr_model_path(),
             det_use_cuda=self.use_cuda,
-            det_use_dml=self.use_dml,
+            det_use_dml=False,
         )
         self.names = {0: "table_text"}
         self.lock = threading.Lock()
@@ -227,12 +227,12 @@ class RapidOCRModel:
 
     def handle_document(
         self,
-        pages: list[babeldoc.document_il.il_version_1.Page],
+        pages: list[babeldoc.format.pdf.document_il.il_version_1.Page],
         mupdf_doc: pymupdf.Document,
         translate_config,
         save_debug_image,
     ) -> Generator[
-        tuple[babeldoc.document_il.il_version_1.Page, YoloResult], None, None
+        tuple[babeldoc.format.pdf.document_il.il_version_1.Page, YoloResult], None, None
     ]:
         for page in pages:
             translate_config.raise_if_cancelled()

{babeldoc-0.3.71/babeldoc → babeldoc-0.4.0/babeldoc/format/pdf}/converter.py

@@ -5,7 +5,7 @@ import unicodedata
 import numpy as np
 from pymupdf import Font
 
-from babeldoc.document_il.frontend.il_creater import ILCreater
+from babeldoc.format.pdf.document_il.frontend.il_creater import ILCreater
 from babeldoc.pdfminer.converter import PDFConverter
 from babeldoc.pdfminer.layout import LTChar
 from babeldoc.pdfminer.layout import LTComponent

babeldoc-0.4.0/babeldoc/format/pdf/document_il/__init__.py

@@ -0,0 +1,47 @@
+from babeldoc.format.pdf.document_il.il_version_1 import BaseOperations
+from babeldoc.format.pdf.document_il.il_version_1 import Box
+from babeldoc.format.pdf.document_il.il_version_1 import Cropbox
+from babeldoc.format.pdf.document_il.il_version_1 import Document
+from babeldoc.format.pdf.document_il.il_version_1 import GraphicState
+from babeldoc.format.pdf.document_il.il_version_1 import Mediabox
+from babeldoc.format.pdf.document_il.il_version_1 import Page
+from babeldoc.format.pdf.document_il.il_version_1 import PageLayout
+from babeldoc.format.pdf.document_il.il_version_1 import PdfCharacter
+from babeldoc.format.pdf.document_il.il_version_1 import PdfFigure
+from babeldoc.format.pdf.document_il.il_version_1 import PdfFont
+from babeldoc.format.pdf.document_il.il_version_1 import PdfFontCharBoundingBox
+from babeldoc.format.pdf.document_il.il_version_1 import PdfFormula
+from babeldoc.format.pdf.document_il.il_version_1 import PdfLine
+from babeldoc.format.pdf.document_il.il_version_1 import PdfParagraph
+from babeldoc.format.pdf.document_il.il_version_1 import PdfParagraphComposition
+from babeldoc.format.pdf.document_il.il_version_1 import PdfRectangle
+from babeldoc.format.pdf.document_il.il_version_1 import PdfSameStyleCharacters
+from babeldoc.format.pdf.document_il.il_version_1 import PdfSameStyleUnicodeCharacters
+from babeldoc.format.pdf.document_il.il_version_1 import PdfStyle
+from babeldoc.format.pdf.document_il.il_version_1 import PdfXobject
+from babeldoc.format.pdf.document_il.il_version_1 import VisualBbox
+
+__all__ = [
+    "BaseOperations",
+    "Box",
+    "Cropbox",
+    "Document",
+    "GraphicState",
+    "Mediabox",
+    "Page",
+    "PageLayout",
+    "PdfCharacter",
+    "PdfFigure",
+    "PdfFont",
+    "PdfFontCharBoundingBox",
+    "PdfFormula",
+    "PdfLine",
+    "PdfParagraph",
+    "PdfParagraphComposition",
+    "PdfRectangle",
+    "PdfSameStyleCharacters",
+    "PdfSameStyleUnicodeCharacters",
+    "PdfStyle",
+    "PdfXobject",
+    "VisualBbox",
+]

{babeldoc-0.3.71/babeldoc → babeldoc-0.4.0/babeldoc/format/pdf}/document_il/backend/pdf_creater.py

@@ -13,12 +13,12 @@ import pymupdf
 from bitstring import BitStream
 
 from babeldoc.assets.embedding_assets_metadata import FONT_NAMES
-from babeldoc.document_il import il_version_1
-from babeldoc.document_il.utils.fontmap import FontMapper
-from babeldoc.document_il.utils.zstd_helper import zstd_decompress
-from babeldoc.translation_config import TranslateResult
-from babeldoc.translation_config import TranslationConfig
-from babeldoc.translation_config import WatermarkOutputMode
+from babeldoc.format.pdf.document_il import il_version_1
+from babeldoc.format.pdf.document_il.utils.fontmap import FontMapper
+from babeldoc.format.pdf.document_il.utils.zstd_helper import zstd_decompress
+from babeldoc.format.pdf.translation_config import TranslateResult
+from babeldoc.format.pdf.translation_config import TranslationConfig
+from babeldoc.format.pdf.translation_config import WatermarkOutputMode
 
 logger = logging.getLogger(__name__)
 
@@ -331,7 +331,7 @@ class PDFCreater:
         self,
         draw_op: BitStream,
         rectangle: il_version_1.PdfRectangle,
-        line_width: float = 1,
+        line_width: float = 0.5,
     ):
         """Draw a rectangle in PDF for visualization purposes.
 
@@ -353,6 +353,8 @@ class PDFCreater:
         draw_op.append(
             rectangle.graphic_state.passthrough_per_char_instruction.encode(),
         )  # Green stroke
+        if rectangle.line_width is not None:
+            line_width = rectangle.line_width
         if line_width > 0:
             draw_op.append(f" {line_width} w ".encode())  # Line width
         draw_op.append(f"{x1} {y1} {width} {height} re ".encode())
@@ -1014,6 +1016,19 @@ class PDFCreater:
                     )
                     translation_config.raise_if_cancelled()
                     original_pdf = pymupdf.open(self.original_pdf_path)
+
+                    if translation_config.debug:
+                        translation_config.raise_if_cancelled()
+                        try:
+                            original_pdf = self.write_debug_info(
+                                original_pdf, translation_config
+                            )
+                        except Exception:
+                            logger.warning(
+                                "Failed to write debug info to dual PDF",
+                                exc_info=True,
+                            )
+
                     if (
                         self.translation_config.only_include_translated_page
                         and should_removed_page
@@ -1043,16 +1058,6 @@ class PDFCreater:
                             translation_config,
                         )
 
-                    if translation_config.debug:
-                        translation_config.raise_if_cancelled()
-                        try:
-                            dual = self.write_debug_info(dual, translation_config)
-                        except Exception:
-                            logger.warning(
-                                "Failed to write debug info to dual PDF",
-                                exc_info=True,
-                            )
-
                     self.save_pdf_with_timeout(
                         dual,
                         dual_out_path,