import streamlit as st
from PIL import Image
import numpy as np
import cv2
from huggingface_hub import from_pretrained_keras

st.header("X-ray tooth segmentation")

st.subheader("Demo improvement Iteration Process")

st.markdown(
    """
This model was created by [SerdarHelli](https://huggingface.co/SerdarHelli/Segmentation-of-Teeth-in-Panoramic-X-ray-Image-Using-U-Net).
"""
)

## Select and load the model
model_id = "SerdarHelli/Segmentation-of-Teeth-in-Panoramic-X-ray-Image-Using-U-Net"
model = from_pretrained_keras(model_id)

## Allows the user to upload an image
archivo_imagen = st.file_uploader("Load her you Image.", type=["png", "jpg", "jpeg"])

## If an image has more than one channel then it is converted to grayscale (1 channel)
def convertir_one_channel(img):
    if len(img.shape) > 2:
        img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
        return img
    else:
        return img


def convertir_rgb(img):
    if len(img.shape) == 2:
        img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
        return img
    else:
        return img


## We'll manipulate the interface so we can use example images
## If the user clicks on an example then the model will run with it
ejemplos = ["dientes_1.png", "dientes_2.png", "dientes_3.png"]


## Let's create three columns; In each one there will be an example image
col1, col2, col3 = st.columns(3)
with col1:
    ## Load the image & show the interface 
    ex = Image.open(ejemplos[0])
    st.image(ex, width=200)
    ## If push the button then, let's use that example within the model
    if st.button("Run this sample 1"):
        archivo_imagen = ejemplos[0]

with col2:
    ex1 = Image.open(ejemplos[1])
    st.image(ex1, width=200)
    if st.button("Run sample 2"):
        archivo_imagen = ejemplos[1]

with col3:
    ex2 = Image.open(ejemplos[2])
    st.image(ex2, width=200)
    if st.button("Run sample 3"):
        archivo_imagen = ejemplos[2]


## If we have an image to input into the model then
## we process it and enter the model
if archivo_imagen is not None:
    ## We load the image with PIL, display it and convert it to a NumPy array
    img = Image.open(archivo_imagen)
    st.image(img, width=850)
    img = np.asarray(img)

    ## We process the image to enter it into the model
    img_cv = convertir_one_channel(img)
    img_cv = cv2.resize(img_cv, (512, 512), interpolation=cv2.INTER_LANCZOS4)
    img_cv = np.float32(img_cv / 255)
    img_cv = np.reshape(img_cv, (1, 512, 512, 1))

    ## We enter the NumPy array to the model
    predicted = model.predict(img_cv)
    predicted = predicted[0]

    ## We return the image to its original shape and add the segmentation masks
    predicted = cv2.resize(
        predicted, (img.shape[1], img.shape[0]), interpolation=cv2.INTER_LANCZOS4
    )
    mask = np.uint8(predicted * 255)  #
    _, mask = cv2.threshold(
        mask, thresh=0, maxval=255, type=cv2.THRESH_BINARY + cv2.THRESH_OTSU
    )
    kernel = np.ones((5, 5), dtype=np.float32)
    mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel, iterations=1)
    mask = cv2.morphologyEx(mask, cv2.MORPH_CLOSE, kernel, iterations=1)
    cnts, hieararch = cv2.findContours(mask, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
    output = cv2.drawContours(convertir_one_channel(img), cnts, -1, (255, 0, 0), 3)

    ## If we successfully got a result then we show it in the interface
    if output is not None:
        st.subheader("Segmentación:")
        st.write(output.shape)
        st.image(output, width=850)
        
        
st.markdown("Thanks for using this application. See you soon, Mate!")