Update app.py

app.py

@@ -9,16 +9,15 @@ model = AutoModel.from_pretrained('jinaai/jina-clip-v1', trust_remote_code=True)
 def compute_similarity(input1_type, input1_text, input1_image, input2_type, input2_text, input2_image):
     """Computes similarity for Text-Text, Image-Image, or Text-Image comparisons."""
 
-    # Ensure images are valid (Gradio may pass `False` instead of `None`)
+    # Ensure images are valid (Gradio sometimes sends `False` instead of `None`)
     input1_image = None if isinstance(input1_image, bool) else input1_image
     input2_image = None if isinstance(input2_image, bool) else input2_image
 
-    # Handle empty inputs properly
+    # Validate inputs
     if input1_type == "Text" and not input1_text.strip():
         return "Error: Input 1 (Text) is empty!"
     if input1_type == "Image" and input1_image is None:
         return "Error: Please upload a valid image for Input 1!"
-
     if input2_type == "Text" and not input2_text.strip():
         return "Error: Input 2 (Text) is empty!"
     if input2_type == "Image" and input2_image is None:
@@ -30,18 +29,15 @@ def compute_similarity(input1_type, input1_text, input1_image, input2_type, inpu
                 emb1 = model.encode_text([input1_text])
                 emb2 = model.encode_text([input2_text])
             elif input1_type == "Image" and input2_type == "Image":
-                emb1 = model.encode_image([Image.fromarray(input1_image)]) if input1_image is not None else None
-                emb2 = model.encode_image([Image.fromarray(input2_image)]) if input2_image is not None else None
+                emb1 = model.encode_image([Image.fromarray(input1_image.astype("uint8"))])
+                emb2 = model.encode_image([Image.fromarray(input2_image.astype("uint8"))])
             else:
                 if input1_type == "Image":
-                    emb1 = model.encode_image([Image.fromarray(input1_image)]) if input1_image is not None else None
+                    emb1 = model.encode_image([Image.fromarray(input1_image.astype("uint8"))])
                     emb2 = model.encode_text([input2_text])
                 else:
                     emb1 = model.encode_text([input1_text])
-                    emb2 = model.encode_image([Image.fromarray(input2_image)]) if input2_image is not None else None
-
-            if emb1 is None or emb2 is None:
-                return "Error: Failed to process one or both inputs."
+                    emb2 = model.encode_image([Image.fromarray(input2_image.astype("uint8"))])
 
             similarity_score = (emb1 @ emb2.T).item()
         return similarity_score