Update app.py

app.py

@@ -7,23 +7,36 @@ import numpy as np
 # Load JinaAI CLIP model
 model = AutoModel.from_pretrained('jinaai/jina-clip-v1', trust_remote_code=True)
 
-def compute_similarity(input1, input2):
+def compute_similarity(input1_type, input1_text, input1_image, input2_type, input2_text, input2_image):
     """
     Computes similarity between:
-    - Image and Text
-    - Image and Image
-    - Text and Text
+    - Text-Text
+    - Image-Image
+    - Text-Image & Image-Text
     """
-    
-    # Detect input types
-    input1_is_text = isinstance(input1, str) and input1.strip() != ""
-    input2_is_text = isinstance(input2, str) and input2.strip() != ""
-    input1_is_image = isinstance(input1, np.ndarray)
-    input2_is_image = isinstance(input2, np.ndarray)
+
+    # Determine input types
+    if input1_type == "Text":
+        input1 = input1_text.strip()
+        input1_is_text = bool(input1)
+        input1_is_image = False
+    else:
+        input1 = input1_image
+        input1_is_text = False
+        input1_is_image = input1 is not None
+
+    if input2_type == "Text":
+        input2 = input2_text.strip()
+        input2_is_text = bool(input2)
+        input2_is_image = False
+    else:
+        input2 = input2_image
+        input2_is_text = False
+        input2_is_image = input2 is not None
 
     # Ensure valid input
     if not (input1_is_text or input1_is_image) or not (input2_is_text or input2_is_image):
-        return "Error: Both inputs must be valid (image or text)!"
+        return "Error: Please provide valid inputs (text or image) for both fields!"
 
     try:
         with torch.no_grad():
@@ -59,32 +72,34 @@ def compute_similarity(input1, input2):
         return f"Error: {str(e)}"
 
 # Gradio UI
-demo = gr.Interface(
-    fn=compute_similarity,
-    inputs=[
-        gr.Radio(["Text", "Image"], label="Input 1 Type", value="Text"),
-        gr.Textbox(label="Text Input 1", visible=True),
-        gr.Image(type="numpy", label="Image Input 1", visible=False),
-
-        gr.Radio(["Text", "Image"], label="Input 2 Type", value="Text"),
-        gr.Textbox(label="Text Input 2", visible=True),
-        gr.Image(type="numpy", label="Image Input 2", visible=False),
-    ],
-    outputs=gr.Textbox(label="Similarity Score / Error", interactive=False),
-    title="JinaAI CLIP Multimodal Similarity",
-    description="Compare similarity between two inputs (Text, Image, or both)."
-)
-
-# Update visibility dynamically
-def update_visibility(input1_type, input2_type):
-    return (
-        input1_type == "Text",  # Text input 1 visibility
-        input1_type == "Image", # Image input 1 visibility
-        input2_type == "Text",  # Text input 2 visibility
-        input2_type == "Image"  # Image input 2 visibility
-    )
-
-# Add event handlers for input type change
-demo.load(update_visibility, inputs=["Input 1 Type", "Input 2 Type"], outputs=["Text Input 1", "Image Input 1", "Text Input 2", "Image Input 2"])
+with gr.Blocks() as demo:
+    gr.Markdown("# JinaAI CLIP Multimodal Similarity")
+    gr.Markdown("Compare similarity between two inputs: **Text-Text, Image-Image, or Image-Text**.")
+
+    with gr.Row():
+        input1_type = gr.Radio(["Text", "Image"], label="Input 1 Type", value="Text")
+        input2_type = gr.Radio(["Text", "Image"], label="Input 2 Type", value="Text")
+
+    input1_text = gr.Textbox(label="Input 1 (Text)", visible=True)
+    input1_image = gr.Image(type="numpy", label="Input 1 (Image)", visible=False)
+
+    input2_text = gr.Textbox(label="Input 2 (Text)", visible=True)
+    input2_image = gr.Image(type="numpy", label="Input 2 (Image)", visible=False)
+
+    output = gr.Textbox(label="Similarity Score / Error", interactive=False)
+
+    def update_visibility(input1_type, input2_type):
+        return (
+            input1_type == "Text",  # Input 1 text visibility
+            input1_type == "Image", # Input 1 image visibility
+            input2_type == "Text",  # Input 2 text visibility
+            input2_type == "Image"  # Input 2 image visibility
+        )
+
+    input1_type.change(update_visibility, inputs=[input1_type, input2_type], outputs=[input1_text, input1_image, input2_text, input2_image])
+    input2_type.change(update_visibility, inputs=[input1_type, input2_type], outputs=[input1_text, input1_image, input2_text, input2_image])
+
+    compute_button = gr.Button("Compute Similarity")
+    compute_button.click(compute_similarity, inputs=[input1_type, input1_text, input1_image, input2_type, input2_text, input2_image], outputs=output)
 
 demo.launch()