app_streamlit.py

import streamlit as st 
import openvino as ov
from pathlib import Path
import numpy as np
from PIL import Image

from src.config import (
    DICT_DIR,
    IMAGE_TYPES,
    IMAGE_EXAMPLE,
    MODEL_DIR,
    DEVICE,
)

from src.image_processing import recognize

# hide deprication warnings which directly don't affect the working of the application
import warnings
warnings.filterwarnings("ignore")

# Setting custom Page Title and Icon with changed layout and sidebar state
st.set_page_config(
    page_title="日本語手書き認識",
    page_icon="",
    layout="centered",
    initial_sidebar_state="auto",
)

# hide the part of the code, as this is just for adding some custom CSS styling but not a part of the main idea 
hide_streamlit_style = """
	<style>
  #MainMenu {visibility: hidden;}
	footer {visibility: hidden;}
  </style>
"""
st.markdown(hide_streamlit_style, unsafe_allow_html=True) # hide the CSS code from the screen as they are embedded in markdown text. Also, allow streamlit to unsafely process as HTML


@st.cache_resource
def init():
    # Load models
    core = ov.Core()
    model = core.read_model(model=Path(MODEL_DIR))
    print("[INFO] Loaded recognition model")

    # Select device (CPU or GPU)
    compiled_model = core.compile_model(model=model, device_name=DEVICE)

    # Fetch Information About Input and Output Layers
    recognition_input_layer = compiled_model.input(0)
    recognition_output_layer = compiled_model.output(0)
    
    print("[INFO] Fetched recognition model")

    # In JA model, there should be blank symbol added at index 0 of each charlist.
    blank_char = "~"

    with Path(DICT_DIR).open(mode="r", encoding="utf-8") as charlist:
        letters = blank_char + "".join(line.strip() for line in charlist)
    print("[INFO] Loaded dictionary")


    return [compiled_model, recognition_input_layer, recognition_output_layer, letters]

def display_text(bounds):
    text = []
    for x in bounds:
        t = x[1]
        text.append(t)
    text = ' '.join(text)
    return text 

HWRmodel = init()
st.set_option('deprecation.showfileUploaderEncoding',False)
st.title('日本語手書き認識')
st.subheader('Tokyo Teachies (DEMO)')
st.subheader('注意：画像には1行のテキストしか含まれていません。')
#st.text('Select source Language from the Sidebar.')

image_file = st.file_uploader("画像をアップロード…",type=IMAGE_TYPES)
if image_file is not None:
    st.subheader('アップロードした画像')
    st.image(image_file,width=450)
else:
    st.subheader('例：')
    image_file = IMAGE_EXAMPLE
    st.image(image_file, use_column_width=450)


if st.button("Recognize"):
    if image_file is not None:
        img = Image.open(image_file).convert('L')
        img = np.array(img)
        
        with st.spinner('テキストの抽出...'):
            recognized_text = recognize(img, HWRmodel[0], HWRmodel[1], HWRmodel[2], HWRmodel[3])
        #st.subheader('Extracted text is ...')
        #text = display_text(recognized_text)
        st.write("".join(recognized_text))