Update app.py

app.py

@@ -1,65 +1,49 @@
-import clip
 import numpy as np
 import torch
 import gradio as gr
 from PIL import Image
 import os
 
+
 # Load the CLIP model
 model, preprocess = clip.load("ViT-B/32")
 device = "cuda" if torch.cuda.is_available() else "cpu"
 model.to(device).eval()
-print(device)
 
 # Define the Business Listing variable
 Business_Listing = "Air Guide"
 
-def find_similarity(image1, image2, text_input):
-    if image1 is None or image2 is None:
-        return None
-
-    image_features = []
-
-    # Preprocess and encode the two images
-    for image in [image1, image2]:
-        image = preprocess(image).unsqueeze(0).to(device)
-        with torch.no_grad():
-            image_feature = model.encode_image(image).float()
-            image_features.append(image_feature)
-
+def find_similarity(image, text_input):
+    # Preprocess the uploaded image
+    image = preprocess(image).unsqueeze(0).to(device)
+    
     # Prepare input text
     text_tokens = clip.tokenize([text_input]).to(device)
-    text_features = model.encode_text(text_tokens).float()
 
-    # Normalize text features
-    text_features /= text_features.norm(dim=-1, keepdim=True)
+    # Encode image and text features
+
 
-    similarities = []
+    with torch.no_grad():
+        image_features = model.encode_image(image).float()
+        text_features = model.encode_text(text_tokens).float()
 
-    # Calculate cosine similarity for each image
-    for image_feature in image_features:
-        image_feature /= image_feature.norm(dim=-1, keepdim=True)
-        similarity = (text_features @ image_feature.T).cpu().detach().numpy()
-        similarities.append(similarity[0, 0])
+    # Normalize features and calculate similarity
+    image_features /= image_features.norm(dim=-1, keepdim=True)
+    text_features /= text_features.norm(dim=-1, keepdim=True)
+    similarity = (text_features @ image_features.T).cpu().numpy()
 
-    # Determine which image has a higher similarity to the text
-    best_match_index = 0 if similarities[0] > similarities[1] else 1
+    return similarity[0, 0]
 
-    return similarities, best_match_index
 
 # Define a Gradio interface
 iface = gr.Interface(
     fn=find_similarity,
-    inputs=[
-        gr.Image(type="pil", label="Image 1"),
-        gr.Image(type="pil", label="Image 2"),
-        "text"
-    ],
-    outputs=["text", "number"],
+    inputs=[gr.Image(type="pil"), "text"],
+    outputs="number",
     live=True,
     interpretation="default",
     title="CLIP Model Image-Text Cosine Similarity",
-    description="Upload two images and enter text to find their cosine similarity.",
+    description="Upload an image and enter text to find their cosine similarity.",
 )
 
-iface.launch()
+iface.launch()