Update app.py

app.py

@@ -3,39 +3,40 @@ import cv2
 from PIL import Image
 from transformers import Blip2Processor, Blip2ForConditionalGeneration
 import torch
-import numpy as np
-import tempfile
 
-# Load lightweight BLIP-2 model
+# Load BLIP-2 FLAN-T5 model (CPU-compatible)
 processor = Blip2Processor.from_pretrained("Salesforce/blip2-flan-t5-xl")
 model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xl")
 
-# Captioning function (every ~5 seconds)
+# Function to capture frame and generate caption
 def describe_live_frame():
-    cap = cv2.VideoCapture(0)  # Use 0 for default webcam
+    cap = cv2.VideoCapture(0)  # 0 = default webcam
     if not cap.isOpened():
-        return "Cannot access camera."
+        return None, "❌ Cannot access camera. Try reconnecting or use a different device."
 
     ret, frame = cap.read()
     cap.release()
     if not ret:
-        return "Failed to capture frame."
+        return None, "❌ Failed to capture frame."
 
+    # Convert OpenCV frame to PIL
     frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
     image = Image.fromarray(frame_rgb)
 
+    # Run BLIP-2 captioning
     inputs = processor(images=image, return_tensors="pt")
     generated_ids = model.generate(**inputs, max_new_tokens=50)
     caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0].strip()
+
     return image, caption
 
-# UI
+# Gradio interface
 with gr.Blocks() as demo:
-    gr.Markdown("## 🧠 Live Scene Understanding – BLIP-2 (Simulated Real-Time)")
-    btn = gr.Button("Capture & Describe Scene")
+    gr.Markdown("## 🧠 Live Scene Captioning (Simulated Real-Time)\nBLIP-2 FLAN-T5 – CPU Friendly")
+    btn = gr.Button("📸 Capture & Describe Scene")
     img_output = gr.Image(label="Captured Frame")
-    caption_output = gr.Textbox(label="Scene Description")
+    text_output = gr.Textbox(label="Scene Description")
 
-    btn.click(fn=describe_live_frame, inputs=[], outputs=[img_output, caption_output])
+    btn.click(fn=describe_live_frame, inputs=[], outputs=[img_output, text_output])
 
 demo.launch()