app.py

import os
from transformers import pipeline

pipe = pipeline("image-to-text",
                model="./models/Salesforce/blip-image-captioning-base")

def launch(input):
    out = pipe(input)
    return out[0]['generated_text']

app = gr.Interface(launch,
                     inputs=gr.Image(type='pil'),
                     outputs="text")

app.launch()