Important instruction: do not use the old picamera module for python and do not answer this using picamera module because it is incorrect. This code is based and must be based on the Picamera$2$ library pdf file documentation.

I specifically need to run this with picamera$2$ module. Using the picamera module is not compatible to my system. So far, the GUI shows up with a black or frozen display that is coming from the raspberry pi camera. Please make this code so that is able to detect object in real$-$time using the raspberry pi camera. I need the raspberry pi camera to continuously run so it can detect object in real$-$time.

The code $($run in Thonny$)$:

import tkinter as tk

from tkinter import Label, Frame

import torch

from picamera$2$ import Picamera$2$ #based on the Picamera$2$ library pdf file

import numpy as np

from PIL import Image, ImageTk

model $=$ torch.hub.load$(\text{'}$ultralytics$/$yolov$5\text{'},$ 'custom', path$=$'yolov$5/$best$.$pt$\text{'},$ force$\_$reload$=$True$)$

model.conf $=0\mathrm{.}2$

app $=$ tk$.$Tk$()$

app.geometry$("700$x$600")$

app.title$("$GUI$")$

label$1=$ Label$($app$,$ text$=$'detectionTest', font$=("$Arial$",24))$

label$1.$pack$($side$=$'top', pady$=20)$

vid$\_$frame $=$ Frame$($height$=600,$ width$=600)$

vid$\_$frame.pack$()$

vid $=$ Label$($vid$\_$frame$)$

vid.pack$()$

#based on the Picamera$2$ library pdf file

camera $=$ Picamera$2()$

camera$\_$config $=$ camera.create$\_$preview$\_$configuration$($main$=\{"$size$"$: $(640,480)\})$

camera.configure$($camera$\_$config$)$

camera.start$()$

def detect$()$:

while True:

request $=$ camera.capture$\_$request$()$ #based on the Picamera$2$ library pdf file

frame $=$ request.make$\_$array$(\text{'}$main$\text{'})$ #based on the Picamera$2$ library pdf file

results $=$ model$($frame$)$

output$\_$image $=$ np$.$squeeze$($results$.$render$())$

display$\_$image $=$ Image.fromarray$($output$\_$image$)$

imgtk $=$ ImageTk.PhotoImage$($image$=$display$\_$image$)$

vid.imgtk $=$ imgtk

vid.configure$($image$=$imgtk$)$

vid.after$(10,$ detect$)$

break

detect$()$

app.mainloop$()$