feat(fontmap): Enhance character glyph detection + Refactor(layout): Move layout parser to a separate class

yadt/document_il/frontend/il_creater.py

@@ -160,8 +160,6 @@ def on_page_number(self, page_number: int):
         assert page_number >= 0
         self.current_page.page_number = page_number
 
-        self.on_page_layout(page_number)
-
     def on_page_base_operation(self, operation: str):
         self.current_page.base_operations = il_version_1.BaseOperations(
             value=operation
@@ -262,6 +260,8 @@ def on_lt_char(self, char: LTChar):
 
         char_id = char.cid
         char_unicode = char.get_text()
+        if '(cid:' not in char_unicode and len(char_unicode) > 1:
+            return
         advance = char.adv
         if char.matrix[0] == 0 and char.matrix[3] == 0:
             vertical = True
@@ -283,39 +283,6 @@ def on_lt_char(self, char: LTChar):
         )
         self.current_page.pdf_character.append(pdf_char)
 
-    def on_page_layout(self, page_number):
-        if (
-            self.translation_config.should_translate_page(page_number + 1)
-            is False
-        ):
-            return
-        pix = self.mupdf[page_number].get_pixmap()
-        image = np.fromstring(pix.samples, np.uint8).reshape(
-            pix.height, pix.width, 3
-        )[:, :, ::-1]
-        h, w = pix.height, pix.width
-        layouts = self.model.predict(image, imgsz=int(pix.height / 32) * 32)[0]
-        id = 0
-        for layout in layouts.boxes:
-            id += 1
-            # Convert the coordinate system from the picture coordinate system to the il coordinate system
-            x0, y0, x1, y1 = layout.xyxy
-            x0, y0, x1, y1 = (
-                np.clip(int(x0 - 1), 0, w - 1),
-                np.clip(int(h - y1 - 1), 0, h - 1),
-                np.clip(int(x1 + 1), 0, w - 1),
-                np.clip(int(h - y0 + 1), 0, h - 1),
-            )
-            page_layout = il_version_1.PageLayout(
-                id=id,
-                box=il_version_1.Box(
-                    x0.item(), y0.item(), x1.item(), y1.item()
-                ),
-                conf=layout.conf.item(),
-                class_name=layouts.names[layout.cls],
-            )
-            self.current_page.page_layout.append(page_layout)
-
     def create_il(self):
         pages = [
             page

yadt/document_il/midend/layout_parser.py

@@ -0,0 +1,62 @@
+import numpy as np
+from pymupdf import Document
+from yadt.document_il import il_version_1
+from yadt.translation_config import TranslationConfig
+from yadt.doclayout import DocLayoutModel
+import logging
+
+logger = logging.getLogger(__name__)
+
+
+class LayoutParser:
+    stage_name = "解析页面布局"
+
+    def __init__(self, translation_config: TranslationConfig):
+        self.translation_config = translation_config
+        self.model = translation_config.doc_layout_model
+        self.progress = None
+
+    def process(self, docs: il_version_1.Document, mupdf_doc: Document):
+        """Generate layouts for all pages that need to be translated."""
+        total = sum(1 for page in docs.page if self.translation_config.should_translate_page(
+            page.page_number + 1))
+        self.progress = self.translation_config.progress_monitor.stage_start(
+            self.stage_name, total)
+
+        for page in docs.page:
+            if not self.translation_config.should_translate_page(page.page_number + 1):
+                continue
+
+            page_number = page.page_number
+            pix = mupdf_doc[page_number].get_pixmap()
+            image = np.fromstring(pix.samples, np.uint8).reshape(
+                pix.height, pix.width, 3
+            )[:, :, ::-1]
+            h, w = pix.height, pix.width
+            layouts = self.model.predict(
+                image, imgsz=int(pix.height / 32) * 32)[0]
+
+            page_layouts = []
+            for layout in layouts.boxes:
+                # Convert the coordinate system from the picture coordinate system to the il coordinate system
+                x0, y0, x1, y1 = layout.xyxy
+                x0, y0, x1, y1 = (
+                    np.clip(int(x0 - 1), 0, w - 1),
+                    np.clip(int(h - y1 - 1), 0, h - 1),
+                    np.clip(int(x1 + 1), 0, w - 1),
+                    np.clip(int(h - y0 + 1), 0, h - 1),
+                )
+                page_layout = il_version_1.PageLayout(
+                    id=len(page_layouts) + 1,
+                    box=il_version_1.Box(
+                        x0.item(), y0.item(), x1.item(), y1.item()
+                    ),
+                    conf=layout.conf.item(),
+                    class_name=layouts.names[layout.cls],
+                )
+                page_layouts.append(page_layout)
+
+            page.page_layout = page_layouts
+            self.progress.advance(1)
+
+        return docs

yadt/document_il/utils/fontmap.py

@@ -44,6 +44,8 @@ def __init__(self, translation_config: TranslationConfig):
         self.fontid2font["fallback"] = self.fallback_font
         self.fontid2font["kai"] = self.kai_font
     def has_char(self, char_unicode: str):
+        if len(char_unicode) != 1:
+            return False
         current_char = ord(char_unicode)
         for font in self.fonts.values():
             if font.has_glyph(current_char):

yadt/high_level.py

@@ -28,6 +28,7 @@
 from yadt.translation_config import TranslationConfig
 from yadt.progress_monitor import ProgressMonitor
 from yadt.document_il.utils.fontmap import FontMapper
+from yadt.document_il.midend.layout_parser import LayoutParser
 import logging
 
 logger = logging.getLogger(__name__)
@@ -155,6 +156,7 @@ def translate(translation_config: TranslationConfig):
         translation_config,
         [
             ILCreater.stage_name,
+            LayoutParser.stage_name,
             ParagraphFinder.stage_name,
             StylesAndFormulas.stage_name,
             ILTranslator.stage_name,
@@ -209,6 +211,16 @@ def translate(translation_config: TranslationConfig):
                 docs, translation_config.get_working_file_path("create_il.debug.json")
             )
 
+        # Generate layouts for all pages
+        logger.debug("start generating layouts")
+        docs = LayoutParser(translation_config).process(docs, doc_input)
+        logger.debug("finish generating layouts")
+
+        if translation_config.debug:
+            xml_converter.write_json(
+                docs, translation_config.get_working_file_path("layout_generator.json")
+            )
+
         ParagraphFinder(translation_config).process(docs)
         logger.debug(f"finish paragraph finder from {temp_pdf_path}")
         if translation_config.debug: