app.py

import streamlit as st
import utils
import cv2
import numpy as np
import io
import tempfile
from PIL import Image 
import moviepy.editor as mpy
from camera_input_live import camera_input_live

st.set_page_config(
    page_title="Hello Text Detection",
    page_icon=":sun_with_face:",
    layout="centered",
    initial_sidebar_state="expanded",)

st.title("Hello Text Dection :sun_with_face:")

st.sidebar.header("메뉴뉴")
source_radio = st.sidebar.radio("Select Source", ["IMAGE", "VIDEO", "WEBCAM"])

st.sidebar.header("Confidence")
conf_threshold = float(st.sidebar.slider("Select the Confidence Threshold", 10, 100, 20))/100

input = None 
if source_radio == "IMAGE":
    st.sidebar.header("Upload")
    input = st.sidebar.file_uploader("Choose an image.", type=("jpg", "png"))
    if input is not None:
        uploaded_image = Image.open(input)
        uploaded_image_cv = cv2.cvtColor(np.array(uploaded_image), cv2.COLOR_RGB2BGR)
        boxes, resized_image = utils.predict_image(uploaded_image_cv, conf_threshold = conf_threshold)
        result_image = utils.convert_result_to_image(uploaded_image_cv, resized_image, boxes, conf_labels=False)
        st.image(result_image, channels = "RGB")
        st.markdown(f"<h4 style='color: blue;'><strong>The result of running the AI inference on an image.</strong></h4>", unsafe_allow_html=True)
    else: 
        st.image("data/intel_rnb.jpg")
        st.write("Click on 'Browse Files' in the sidebar to run inference on an image." )

def play_video(video_source):
    camera = cv2.VideoCapture(video_source)
    fps = camera.get(cv2.CAP_PROP_FPS)
    temp_file_2 = tempfile.NamedTemporaryFile(delete=False,suffix='.mp4')
    video_row=[]
    # frame
    total_frames = int(camera.get(cv2.CAP_PROP_FRAME_COUNT))
    progress_bar = st.progress(0)
    frame_count = 0
    st_frame = st.empty()
    while(camera.isOpened()):
        ret, frame = camera.read()
        if ret:
            try:
                boxes, resized_image = utils.predict_image(frame, conf_threshold)
                visualized_image = utils.convert_result_to_image(frame, resized_image, boxes, conf_labels=False)
            except:
                visualized_image = frame
            st_frame.image(visualized_image, channels = "BGR")
            video_row.append(cv2.cvtColor(visualized_image,cv2.COLOR_BGR2RGB))
            frame_count +=1 
            progress_bar.progress(frame_count/total_frames, text=None)
    
        else:
            progress_bar.empty()
            camera.release()
            st_frame.empty()
            break
        clip = mpy.ImageSequenceClip(video_row, fps = fps)
        clip.write_videofile(temp_file_2.name)
        st.video(temp_file_2.name)

# 파일 업로드 처리
temporary_location = None

if source_radio == "VIDEO":
    st.sidebar.header("Upload")
    input_file = st.sidebar.file_uploader("Choose a video.", type=("mp4"))
    if input_file is not None:
        # 파일을 임시 경로에 저장
        g = io.BytesIO(input_file.read())
        temporary_location = "upload.mp4"
        with open(temporary_location, "wb") as out:
            out.write(g.read())
        out.close()
    # 업로드된 비디오 파일이 있는 경우 비디오 재생
    if temporary_location is not None:
        play_video(temporary_location)
    else:
        st.video("data/sample_video.mp4")
        st.write("Click on 'Browse Files' in the sidebar to run inference on a video.")

if source_radio == "WEBCAM":
    input = camera_input_live()
    uploaded_image = Image.open(input)
    uploaded_image_cv = cv2.cvtColor(np.array(uploaded_image), cv2.COLOR_RGB2BGR)
    boxes, resized_image = utils.predict_image(uploaded_image_cv, conf_threshold)
    visualized_image = utils.convert_result_to_image(uploaded_image_cv, resized_image, boxes, conf_labels=False)
    st.image(visualized_image, channels = "RGB")