update multi apps

desktop_env/evaluators/metrics/chrome.py

@@ -143,6 +143,11 @@ def compare_pdfs(pdf1_path: Union[str, List[str]], pdf2_path: Union[str, List[st
 import fitz
 from PIL import Image
 from io import BytesIO
+from borb.pdf import Document
+from borb.pdf import PDF
+
+from pathlib import Path
+import typing
 
 def compare_pdf_images(pdf1_path: str, pdf2_path: str, **kwargs) -> float:
     def extract_images_from_pdf(pdf_path):
@@ -151,24 +156,32 @@ def compare_pdf_images(pdf1_path: str, pdf2_path: str, **kwargs) -> float:
 
         for page_number in range(pdf_document.page_count):
             page = pdf_document[page_number]
-            image_list = page.get_images(full=True)
+            pixmap = page.get_pixmap()
 
-            for img_index, img_info in enumerate(image_list):
-                base_image = pdf_document.extract_image(img_index)
-                image_bytes = base_image["image"]
+            img = Image.frombytes("RGB", [pixmap.width, pixmap.height], pixmap.samples)
 
-                images.append(BytesIO(image_bytes))
+            images.append(img)
 
         return images
     
+    def fix_pdf(in_path: Path, out_path: Path) -> None:
+        doc: typing.Optional[Document] = None
+        with open(in_path, "rb") as fh:
+            doc = PDF.loads(fh)
+        with open(out_path, "wb") as fh:
+            PDF.dumps(fh, doc)
+    
+    fix_pdf(Path(pdf1_path), Path(pdf1_path))
+    fix_pdf(Path(pdf2_path), Path(pdf2_path))
+
     images1 = extract_images_from_pdf(pdf1_path)
     images2 = extract_images_from_pdf(pdf2_path)
 
     if len(images1) != len(images2):
         return 0.
 
-    for i, (img1, img2) in enumerate(zip(images1, images2), 1):
-        if Image.open(img1).tobytes() != Image.open(img2).tobytes():
+    for img1, img2 in zip(images1, images2):
+        if img1.tobytes() != img2.tobytes():
             return 0.
     
     return 1.