app.py

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

# Load your model
processor = BlipProcessor.from_pretrained("mshsahmed/blip-vqa-finetuned-kvasir-v58849")
model = BlipForConditionalGeneration.from_pretrained("mshsahmed/blip-vqa-finetuned-kvasir-v58849")

def vqa_pipeline(image, question):
    inputs = processor(image, question, return_tensors="pt")
    out = model.generate(**inputs)
    answer = processor.decode(out[0], skip_special_tokens=True)
    return answer

iface = gr.Interface(
    fn=vqa_pipeline,
    inputs=[gr.Image(type="pil"), gr.Textbox(lines=1, placeholder="Ask a question...")],
    outputs="text",
    title="Medical VQA Demo",
    description="Upload an image and ask a question. The model will answer based on the image content."
)

iface.launch()