import streamlit as st
from random import randint
#from .session_state import get_session_state

import cv2
import pandas
from PIL import Image
import numpy as np
import tensorflow as tf
from tensorflow.keras.applications.resnet50 import preprocess_input 
from tensorflow.keras.preprocessing.image import img_to_array    

st.title('Palm Identification')
st.markdown("This is a Deep Learning application to identify if a satellite image clip contains Palm trees.\n")
st.markdown('The predicting result will be "Palm", or "Others".')
st.markdown('You can click "Browse files" multiple times until adding all images before generating prediction.\n')

#uploaded_file = st.file_uploader("Upload an image file", type="jpg", accept_multiple_files=True)

#imageContainer = st.empty()

#closeImage = st.button("clear all images")
img_height = 224
img_width = 224
class_names = ['Palm', 'Others']
model = tf.keras.models.load_model('model')

state = st.session_state.get_session_state()
if not state.widget_key:
    state.widget_key = str(randint(1000, 100000000))

uploaded_file = st.file_uploader(
    "Choose a file", accept_multiple_files=True, key=state.widget_key)

if st.button('clear uploaded_file'):
    state.widget_key = str(randint(1000, 100000000))
state.sync()

#Generate_pred = st.button("Generate Prediction")
#with st.form("form", clear_on_submit=True):
#    uploaded_file = st.file_uploader("Upload image files", type="jpg", accept_multiple_files=True)
#    if uploaded_file is not None:
#        st.image(uploaded_file, width=100)
        
#    submitted = st.form_submit_button("Toggle here to predict or to delete the data")
            
    #    if submitted and uploaded_file is not None:
    
#            for file in uploaded_file:
#                img = Image.open(file)
#                img_array = img_to_array(img)
#                img_array = tf.expand_dims(img_array, axis = 0) # Create a batch
#                processed_image = preprocess_input(img_array)
        
#                predictions = model.predict(processed_image)
#                score = predictions[0]
#                st.markdown("Predicted class of the image {} is : {}".format(file, class_names[np.argmax(score)]))
    
#            uploaded_file = None