import os
import io
import cv2
import requests
import numpy as np
import streamlit as st

from PIL import Image
from skimage.io import imread


def infer() -> None:
    st.title("Overhead MNIST classification app")

    # select an input image file
    image_file_buffer = st.sidebar.file_uploader(
        "Select input image", type=["jpg", "jpeg"]
    )
    # read the image
    if image_file_buffer is not None:
        image = Image.open(image_file_buffer)
        image_array = np.array(image)
        st.image(image_array, caption=f"Input image: {image_file_buffer.name}")
    else:
        st.write("Input image: not selected")

    # run inference when the option is invoked by the user
    infer_button = st.sidebar.button("Run inference")
    if infer_button:
        files = {"image_file": (image_file_buffer.name, image_file_buffer.getvalue())}

        # if the deployment is on local machine
        response = requests.post(
            "https://abhishekrs4-overhead-mnist.hf.space/predict",
            files=files,
        )

        # if the deployment is on hugging face
        # response = requests.post(
        #     "http://127.0.0.1:7860/predict",
        #     files=files,
        # )

        st.write("The following is the prediction")
        st.write(response.json())
    return


def app_info() -> None:
    st.title("App info")
    st.markdown("_Task - Overhead MNIST classification_")
    st.markdown(
        "_Project repo - [https://github.com/AbhishekRS4/overhead_mnist](https://github.com/AbhishekRS4/overhead_mnist)_"
    )
    st.markdown(
        "_Dataset - [Overhead MNIST dataset](https://www.kaggle.com/datasets/datamunge/overheadmnist/)_"
    )
    st.header("Brief description of the project")
    st.write(
        "The Overhead MNIST dataset contains images extracted from satellite data."
    )
    st.write(
        "This dataset contains instances for 10 classes --- car, harbor, helicopter, oil_gas_field, parking_lot, plane, runway_mark, ship, stadium, and storage_tank."
    )
    st.write("A custom architecture is modeled for the classification task.")
    st.write("The best performing model has been used for the deployed application.")
    return


app_modes = {
    "App Info": app_info,
    "Overhead MNIST Inference App": infer,
}


def start_app() -> None:
    selected_mode = st.sidebar.selectbox("Select mode", list(app_modes.keys()))
    app_modes[selected_mode]()
    return


def main() -> None:
    start_app()
    return


if __name__ == "__main__":
    main()