fix pdf

app.py

@@ -15,8 +15,10 @@
 # limitations under the License.
 
 import traceback
+import base64
 
 import gradio as gr
+import cv2
 
 from paddlenlp import Taskflow
 from paddlenlp.utils.doc_parser import DocParser
@@ -83,6 +85,12 @@ def dbc2sbc(s):
     return rs
 
 
+def np2base64(image_np):
+    image = cv2.imencode('.jpg', image_np)[1]
+    base64_str = str(base64.b64encode(image))[2:-1]
+    return base64_str
+
+
 def process_path(path):
     error = None
     if path:
@@ -179,8 +187,13 @@ def process_doc(document, schema, ocr_lang, layout_analysis):
     }
     prediction = run_taskflow(document, schema, argument)[0]
 
+    if document.endswith(".pdf"):
+        _image = doc_parser.read_pdf(document)
+    else:
+        _image = doc_parser.read_image(document)
+
     img_show = doc_parser.write_image_with_results(
-        document,
+        np2base64(_image),
         result=prediction,
         return_image=True)
     img_list = [img_show]