app.py

import gradio as gr
from transformers import pipeline

# Define the task and model for the pipeline
task = 'image-to-text'
model = 'Salesforce/blip-image-captioning-base'
caption_pipeline = pipeline(task=task, model=model, max_new_tokens=100)

def get_caption(image):
    # Directly pass the PIL image to the pipeline
    result = caption_pipeline(image)
    return result[0]['generated_text']

# Create a Gradio interface
result = gr.Interface(fn=get_caption, 
                      inputs=[gr.Image(label="Upload Image", type="pil")], 
                      outputs="text",
                      examples=['./images/cow.jpeg', './images/christmas_dog.jpeg']
                      )

# Launch the interface
result.launch(inline=False)