app.py

import streamlit as st
from transformers import AutoModelForCausalLM, AutoTokenizer
from PIL import Image
import requests
from io import BytesIO

model_name = "nlpconnect/vit-gpt2-image-captioning"
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

def generate_caption(image):
    image = image.convert("RGB")
    
    image = image.resize((224, 224))
    
    inputs = tokenizer("Image caption: ", return_tensors="pt", max_length=30, truncation=True)

    with st.spinner("Generating caption..."):
        caption_ids = model.generate(input_ids=inputs["input_ids"], attention_mask=inputs["attention_mask"])

    generated_caption = tokenizer.decode(caption_ids[0], skip_special_tokens=True)
    
    return generated_caption

def main():
    st.title("Image Captioning App")
    
    with st.form("my_form"):
        uploaded_file = st.file_uploader("Choose an image file", type=["jpg", "jpeg", "png"])

        if uploaded_file is not None:
            image = Image.open(uploaded_file)
            st.image(image, caption="Uploaded Image", use_column_width=True)

        clicked = st.form_submit_button("Generate Caption")
        
        if clicked and uploaded_file is not None:
            caption = generate_caption(image)
            st.success("Generated Caption:")
            st.write(caption)

if __name__ == "__main__":
    main()