Room WireFrame recunstruction and corner points detection from Single RGB with Ai -- 2

In order to carry out a research project, we need a Python script that can calculate points in an interior room hall using OpenCV and AI (or any other technology in Python) and determine its coordinates by drawing markers in 2D in the picture.

The points in question are as follows:
1. The corners of the floor
2. A point of the ceiling
3. Accurate calculation of corners that are hidden behind other objects
The main challenge of this project is that in the input images, sometimes a corner of the wall corner may be hidden behind an object and cannot be seen, for example, furniture that blocks the corner from being seen, and the system must guess the exact location of the corner. picture number 3)

The second challenge is that the points must be calculated accurately and without errors (this issue is very important to us).

Finally, the intended output should be an image and a file in json format that lists the coordinates of each point in order from the left side of the scene to the right, and the last one is the ceiling.
In the appendix of the description, there is a number of images that specify the input and output we are looking for

You can also use any project, source or tool to do this, for example the following project is one of the projects that
It solves part of our needs:

1. https://github.com/bertjiazheng/Structured3D

2. https://github.com/vevenom/ScanNet-Layout

3. https://github.com/GitBoSun/roomnet/blob/master/README.md

4 . https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123630715.pdf