import streamlit as st
from transformers import pipeline
from PIL import Image

@st.cache_resource
def get_model_hotdog_classification():
	model = pipeline(task="image-classification", model="julien-c/hotdog-not-hotdog")
	return model

@st.cache_resource
def get_model_image_captioning():
    captioner = pipeline("image-to-text", model="Salesforce/blip-image-captioning-base")
    return captioner

tabs1,tabs2 = st.tabs(['Hot Dog? Or Not?','Imaage Captioning'])
with tabs1:
    st.title("Hot Dog? Or Not?")

    file_name = st.file_uploader("Upload a hot dog candidate image",key="hotdog_image")

    if file_name is not None:
        col1, col2 = st.columns(2)

        image = Image.open(file_name)
        col1.image(image, use_column_width=True)
        model = get_model_hotdog_classification()
        predictions = model(image)

        col2.header("Probabilities")
        for p in predictions:
            col2.subheader(f"{ p['label'] }: { round(p['score'] * 100, 1)}%")

with tabs2:
    st.title("Image Captioning")

    file_name = st.file_uploader("Upload an image to caption",key="caption_image")

    if file_name is not None:
        col1, col2 = st.columns(2)
        image = Image.open(file_name)

        col1.image(image, use_column_width=True)

        captioner = get_model_image_captioning()

        with col2:
            st.header("generated caption")
            with st.spinner("Generating caption..."):
                predictions = captioner(image)
            for generated_text in predictions:
                st.write(f"\n{generated_text['generated_text']}")