streamlit_app.py

import io
from typing import TypedDict, Tuple

import gettext
_ = gettext.gettext

try:
    localizator = gettext.translation('base', localedir='locales', languages=['kr'])
    localizator.install()
    _ = localizator.gettext
except Exception:
    pass

from PIL import Image, ImageDraw

import streamlit as st
st.set_page_config(page_title="modi, 모두의 아이돌", page_icon="🎙️", menu_items={"Get help": None, "Report a bug": "https://forms.gle/kHDsXG9ctMXs75AJ9", "About": "# 모디 modi 🎙️\n[노션 페이지](https://jumto.notion.site/Modi-293e5832633a402f8e2de0278eaee975)"})

import flask_app
import rekognition.utils_alert
import rekognition
import rekognition.utils
import utils_streamlit
import streamlit_config


class MatchParsed(TypedDict):
    idol: rekognition.Idol
    similarity: float


def suggest_line_width(size: Tuple[float, float]) -> int:
    longer = max(size)
    return max(1, int(longer * 0.01))


@rekognition.utils_alert.alert_slack_when_exception(will_raise=False)
def main():
    st.markdown(_('# 지금 보이는 스타는 누구일까요?'))
    file = st.file_uploader(_("알고 싶은 스타가 보이는 이미지를 업로드해주세요"), type=streamlit_config.image_types)

    if file is not None:
        file_bytes = file.read()
        try:
            searched_response: flask_app.UploadPostPayload = utils_streamlit.call_api(url_path='/upload', method='POST', files=dict(file=file_bytes))
        except utils_streamlit.ErrorResponse as e:
            if any(locked_error in e.body['exception'] for locked_error in ['PausedError', 'TooHotError']):
                st.write(_("요청이 너무 많아 일시적으로 잠겨있습니다. 10분 뒤 다시 시도해주세요."))
                return
            else:
                rekognition.utils_alert.alert_slack_exception(e)
                st.write(_("결과를 받는 중에 문제가 발생했습니다. 10초 후 다시 시도해주세요."))
                return
        except Exception as e:
            rekognition.utils_alert.alert_slack_exception(e)
            st.write(_("결과를 받는 중에 문제가 발생했습니다. 10초 후 다시 시도해주세요."))
            return

        if len(searched_response['searcheds']) == 0:
            st.error(_("얼굴을 찾을 수 없습니다."))
        else:
            image = Image.open(io.BytesIO(file_bytes))

            canvas_image = image.copy()
            canvas_image = rekognition.utils.roughly_fit_to(canvas_image, (400, 400), zoom_threshold=1.)
            draw = ImageDraw.Draw(canvas_image)
            for searched_each in searched_response['searcheds']:
                bounding_box_corners = rekognition.utils_boto3.to_abs_bounding_box_corners(bounding_box=searched_each['searched_face_bounding_box'], size=canvas_image.size)
                draw.rounded_rectangle(xy=bounding_box_corners, radius=suggest_line_width(canvas_image.size) * 4, width=suggest_line_width(canvas_image.size))
            st.image(image=canvas_image)

            containers_searched_result = [st.container() for __ in range(len(searched_response['searcheds']))]
            for container, searched_each in zip(containers_searched_result, searched_response['searcheds']):
                column1, column2 = container.columns([1, 2])
                with column1:
                    margined_face_bounding_box = rekognition.utils_boto3.margin_bounding_box(bounding_box=searched_each['searched_face_bounding_box'])
                    margined_face_bounding_box_corners = rekognition.utils_boto3.to_abs_bounding_box_corners(margined_face_bounding_box, size=image.size)
                    face_image = rekognition.utils.roughly_fit_to(image.crop(box=margined_face_bounding_box_corners), (150, 150), zoom_threshold=1.)
                    st.image(face_image)
                with column2:
                    for each_match in searched_each['matches']:
                        try:
                            idol = rekognition.Idol.from_external_image_id(each_match['Face']['ExternalImageId'])
                            similarity = each_match['Similarity']
                        except Exception as e:
                            rekognition.utils_alert.alert_slack_exception(exception=e)
                            st.write(_("에러가 발생했습니다. 개발자에게 연락해주세요."))
                        else:
                            try:
                                detail_response: flask_app.DetailPayload = utils_streamlit.call_api(url_path=f'/detail/{idol.idol_id}')
                                detail_md = detail_response['markdown']
                            except Exception as e:
                                rekognition.utils_alert.alert_slack_exception(exception=e)
                                st.write(idol.idol_id)
                            else:
                                st.markdown(detail_md)
                container.markdown("---")


if __name__ == '__main__':
    if utils_streamlit.ask_admin_password_if_needed():
        main()