app.py

# app.py

import gradio as gr
import torch
from transformers import BlipProcessor, BlipForConditionalGeneration
import cv2
from PIL import Image

# Load BLIP captioning model
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")

device = torch.device("cpu")
model.to(device)

# Live webcam captioning generator
def webcam_caption():
    cap = cv2.VideoCapture(0)  # open webcam
    while True:
        ret, frame = cap.read()
        if not ret:
            break

        # Convert OpenCV frame (BGR) to RGB PIL Image
        frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
        image = Image.fromarray(frame_rgb)

        # Generate caption
        inputs = processor(images=image, return_tensors="pt").to(device)
        out = model.generate(**inputs, max_new_tokens=50)
        caption = processor.decode(out[0], skip_special_tokens=True)

        yield frame_rgb, caption

    cap.release()

# Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("## 🎥 Live Webcam BLIP Captioning (CPU)")
    video = gr.Image(label="Webcam Stream")
    text = gr.Textbox(label="Caption")

    demo.load(
        fn=webcam_caption,
        inputs=None,
        outputs=[video, text],
        every=2  # call generator every 2 sec (adjust if you want)
    )

demo.launch()