Show original

app.py

@@ -94,7 +94,7 @@ def inference(pdf_path, page_num):
     page_image = images[0]
     paddle_ocr = get_paddle_ocr(page_image)
     smoldocling_ocr = get_smoldocling_ocr(page_image)
-    return docling_ocr, paddle_ocr, smoldocling_ocr
+    return page_image, docling_ocr, paddle_ocr, smoldocling_ocr
 
 title = "OCR Arena"
 description = "A simple Gradio interface to extract text from PDFs and compare OCR models"
@@ -118,9 +118,10 @@ with gr.Blocks(theme=gr.themes.Glass()) as demo:
                         clear_btn = gr.ClearButton(components=[pdf, page_num])
                         submit_btn = gr.Button("Submit", variant='primary')
 
-                    submit_btn.click(inference, inputs=[pdf, page_num], outputs=[docling_ocr_out, paddle_ocr_out, smoldocling_ocr_out])
+                    submit_btn.click(inference, inputs=[pdf, page_num], outputs=[original, docling_ocr_out, paddle_ocr_out, smoldocling_ocr_out])
 
         with gr.Column():
+            original = gr.Image(width=320, height=320, label="Original Page")
             docling_ocr_out = gr.Textbox(label="Docling OCR Output", type="text")
             paddle_ocr_out = gr.Textbox(label="Paddle OCR Output", type="text")
             smoldocling_ocr_out = gr.Textbox(label="SmolDocling OCR Output", type="text")