Update app.py

app.py

@@ -232,21 +232,47 @@ def process_file_and_create_pdf(file):
 
 # Gradio Interface
 @GPU
-def process_file_for_gradio(file):
+def process_file_for_gradio(image_file, pdf_file):
     """
-    Wrapper function for Gradio interface.
-    This function calls the main processing logic and returns the outputs
-    in the format required by the gr.Interface.
+    Wrapper function for Gradio interface with separate inputs.
+    This function checks which input was provided and calls the main
+    processing logic accordingly.
     """
-    output_path, input_image = process_file_and_create_pdf(file)
-    return output_path, input_image
+    if image_file is not None:
+        # The gr.Image component returns a PIL Image object
+        # We need to save it to a temporary file for the main function
+        temp_dir = tempfile.mkdtemp()
+        image_path = os.path.join(temp_dir, "uploaded_image.png")
+        image_file.save(image_path)
+        
+        # Create a mock file object to be compatible with the main function
+        class MockFile:
+            def __init__(self, name):
+                self.name = name
+        
+        mock_file = MockFile(image_path)
+        output_path, input_image = process_file_and_create_pdf(mock_file)
+        shutil.rmtree(temp_dir)
+        return output_path, input_image
+    
+    elif pdf_file is not None:
+        # The gr.File component passes a temporary file object directly
+        output_path, input_image = process_file_and_create_pdf(pdf_file)
+        return output_path, input_image
+    
+    else:
+        return None, None
+
 
 demo = gr.Interface(
     fn=process_file_for_gradio,
-    inputs=gr.File(label="Upload an Image (PNG, JPG) or a PDF", file_types=['.png', '.jpg', '.jpeg', '.pdf']),
+    inputs=[
+        gr.Image(label="Upload an Image", type="pil"),
+        gr.File(label="Upload a PDF", file_types=['.pdf'])
+    ],
     outputs=[
         gr.File(label="Download OCR Results PDF", interactive=False, visible=True),
-        gr.Image(label="Uploaded Image Preview", interactive=False)
+        gr.Image(label="Uploaded File Preview", interactive=False)
     ],
     title="OCR App with PaddleOCR and TrOCR",
     description="Upload an image or a multi-page PDF to get an output PDF with the recognized text from each page. The output PDF will be downloaded automatically.",