Update app.py

app.py

@@ -1,58 +1,57 @@
-# 1. Install Required Libraries (run once in terminal or notebook)
-# pip install gradio transformers torch torchvision pillow
-
-# 2. Import Libraries
 import gradio as gr
 from transformers import CLIPProcessor, CLIPModel
 from PIL import Image
 import torch
 
-# 3. Load the Pre-trained CLIP Model
+# Load CLIP model and processor
 model_name = "openai/clip-vit-base-patch16"
 processor = CLIPProcessor.from_pretrained(model_name)
 model = CLIPModel.from_pretrained(model_name)
 
-# 4. Define the Matching Function
+# Define matching function
 def match_image_with_descriptions(image, descriptions):
     if not image or not descriptions.strip():
-        return "Please upload an image and enter at least one description."
+        return {"Error": "Please upload an image and enter descriptions."}
 
-    # Parse user input into a list of captions
+    # Split user input into lines
     captions = [line.strip() for line in descriptions.strip().split('\n') if line.strip()]
     
     if len(captions) < 2:
-        return "Please enter at least two descriptions to compare."
+        return {"Error": "Please enter at least two descriptions."}
 
-    # Preprocess inputs
+    # Tokenize inputs
     inputs = processor(text=captions, images=image, return_tensors="pt", padding=True)
-    
-    # Run CLIP model
+
+    # Run inference
     with torch.no_grad():
         outputs = model(**inputs)
 
-    # Get prediction scores
-    logits_per_image = outputs.logits_per_image  # shape: [1, num_captions]
+    # Get probabilities
+    logits_per_image = outputs.logits_per_image  # [1, num_captions]
     probs = logits_per_image.softmax(dim=1)[0]   # shape: [num_captions]
 
-    # Format results into a dictionary with raw floats (0.0 - 1.0)
-    result_dict = {captions[i]: probs[i].item() for i in range(len(captions))}
+    # Convert to percentage and build result dict
+    result_dict = {
+        captions[i]: round(probs[i].item() * 100, 2)  # convert to percentage
+        for i in range(len(captions))
+    }
 
-    # Pick the best match
-    best_caption = max(result_dict, key=result_dict.get)
+    # Sort from best match to worst
+    sorted_results = dict(sorted(result_dict.items(), key=lambda item: item[1], reverse=True))
 
-    return best_caption, result_dict
+    return sorted_results
 
-# 5. Build Gradio Interface
+# Create the Gradio interface
 iface = gr.Interface(
     fn=match_image_with_descriptions,
     inputs=[
         gr.Image(type="pil", label="Upload an Image"),
         gr.Textbox(lines=6, placeholder="Enter one description per line...", label="Descriptions")
     ],
-    outputs=gr.Label(label="Best Match with Confidence Scores"),
-    title="🧠 CLIP Image-Text Matcher",
-    description="Upload an image and enter multiple captions (one per line). The AI will compare them and show which caption best fits the image.",
+    outputs=gr.JSON(label="Match Scores (Sorted by Confidence %)"),
+    title="🧠 CLIP Image-Text Matcher (Sorted, %)",
+    description="Upload an image and enter multiple captions (one per line). The AI will compare each one and show match scores in % — sorted from best to worst.",
 )
 
-# 6. Launch the App
+# Launch the app
 iface.launch()