import gradio as gr
from model import get_results_model
from model import model_
import cv2

IMAGES = 0


def predict_image(image):
    global IMAGES
    paths = f'images/image_{IMAGES}.jpg'
    cv2.imwrite(paths, image)
    IMAGES += 1
    result = get_results_model(paths, model_)
    if result[2] < 0.001:
        label_img = 'Unrecognised'
        pred_acc = ''
    else:
        label_img = result[1]
        pred_acc = f'Probability: &nbsp; **{(result[2] * 100):.2f} %**'
    return result[0], f'<font size="10"> Class: &nbsp; **{label_img}** &nbsp;&nbsp;&nbsp;&nbsp; {pred_acc}</font>'


with gr.Blocks() as demo:
    gr.Markdown('**<font size="10">MRI Assistant</font>**')
    with gr.Row():
        with gr.Column():
            image_input = gr.Image(label='MRI')
            label = gr.Markdown("")
        image_output = gr.Image(label='AI results')

    image_button = gr.Button("Predict results")

    gr.Markdown(r"""
                <font size="10">Social:</font>\
                &nbsp;&nbsp; <font size="7">*1.*</font>&nbsp;&nbsp; <font size="6"> [*Developers*](https://t.me/HenSolaris) </font>\
                &nbsp;&nbsp; <font size="7">*2.*</font>&nbsp;&nbsp; <font size="6"> [*Telegram bot*](https://t.me/Altsheimer_AI_bot) </font>
                """)

    image_button.click(predict_image, inputs=image_input, outputs=[image_output, label])

demo.launch()

print('launched!')