app.py

import streamlit as st
import streamlit.components.v1 as components
from PIL import Image

from predict import generate_text
from model import load_clip_model, load_gpt_model, load_model


# Configure Streamlit page
st.set_page_config(page_title="Caption Machine", page_icon="💥")

# Set Session

model, image_transform, tokenizer = load_model()

if 'model' not in st.session_state:
    st.session_state['model'] = model

if 'image_transform' not in st.session_state:
    st.session_state['image_transform'] = image_transform

if 'tokenizer' not in st.session_state:
    st.session_state['tokenizer'] = tokenizer



# Force responsive layout for columns also on mobile
st.write(
    """<style>
    [data-testid="column"] {
        width: calc(50% - 1rem);
        flex: 1 1 calc(50% - 1rem);
        min-width: calc(50% - 1rem);
    }
    </style>""",
    unsafe_allow_html=True,
)

# Render Streamlit page
st.title("Image Captioner")
st.markdown(
    "This app generates Image Caption using OpenAI's [GPT-2](https://openai.com/research/better-language-models) and [CLIP](https://openai.com/research/clip) model."
)

# st.subheader("Model Architecture")

# image = Image.open('model.png')

# st.image(image, caption=None, width=500)

upload_file = st.file_uploader("Upload an image:", type=['png','jpg','jpeg'])


# Checking the Format of the page
if upload_file is not None:
    img = Image.open(upload_file)
    st.image(img)
    st.write("Image Uploaded Successfully")

    # gpt_model, tokenizer = load_gpt_model()
    caption = generate_text(st.session_state['model'], img, st.session_state['tokenizer'], st.session_state['image_transform'])

    st.write(caption)