do it normally

app.py

@@ -5,7 +5,7 @@ import os
 import base64
 import spaces
 import io
-import numpy as np
+# import numpy as np
 from PIL import Image
 import io
 
@@ -49,11 +49,13 @@ model = AutoModel.from_pretrained('ucaslcl/GOT-OCR2_0', trust_remote_code=True,
 model = model.eval().cuda()
 model.config.pad_token_id = tokenizer.eos_token_id
 
+
+
 def process_input_image(image):
-    if isinstance(image, np.ndarray):
-        return Image.fromarray(image.astype('uint8'), 'RGB')
-    elif isinstance(image, str):
+    if isinstance(image, str):
         return Image.open(image)
+    elif isinstance(image, Image.Image):
+        return image
     else:
         raise ValueError("Unsupported image type")
 
@@ -63,6 +65,7 @@ def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None):
         if image is None:
             return "No image provided", None
 
+        # Process the input image
         pil_image = process_input_image(image)
 
         with io.BytesIO() as buffer:
@@ -127,7 +130,7 @@ with gr.Blocks() as demo:
     gr.Markdown(description)
     with gr.Row():
         with gr.Column():
-            image_input = gr.Image(type="filepath", label="Input Image")
+            image_input = gr.Image(type="pil", label="Input Image")
             task_dropdown = gr.Dropdown(
                 choices=[
                     "Plain Text OCR",