app.py

import streamlit as st
from PIL import Image

from transformers import GPT2TokenizerFast, ViTImageProcessor, VisionEncoderDecoderModel

# load image captioning model and corresponding tokenizer and image processor
model = VisionEncoderDecoderModel.from_pretrained("jojo-ai-mst/image-vision-cap")
tokenizer = GPT2TokenizerFast.from_pretrained("jojo-ai-mst/image-vision-cap")
image_processor = ViTImageProcessor.from_pretrained("jojo-ai-mst/image-vision-cap")


def generate_caption(image):
    image = Image.open(image)
    pixel_values = image_processor(image, return_tensors="pt").pixel_values

    # autoregressively generate caption (uses greedy decoding by default)
    generated_ids = model.generate(pixel_values)
    generated_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
    print(generated_text)

    return generated_text

st.header("Welcome to Vision Caption Prototype",divider="rainbow")

uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])

st.markdown("""
    <style>
    .stButton button {
        background-color: green;
        color: white;
    }
    </style>
    """, unsafe_allow_html=True)

st.divider()

if st.button("Get Answer"):
    if uploaded_file is not None:
        st.header("Result")
        st.image(uploaded_file, caption=uploaded_file.name, use_column_width=True)

        #answer = "The answer will be generated by AI"

        caption = generate_caption(uploaded_file)

        st.subheader(caption)
    else:
        st.write("Please upload an image and type a question.")