2D to 3D Floorplan Converter
Andre Ahuna
The plan is to build a Unity application, which allows users to upload images of floorplans and constructs a 3D approximation of it.
The initial target is wall detection and position parsing from an image to an easy-to-digest format for the Unity client. Window and door detection is a must as well. All other viable detected classes will be fixed on a later date.
The method of parsing is yet to be determined. Possibly a general LLM, potentially a pre-trained LLM, maybe CV or a combination of all of them.
Resulting app should be usable in VR, allowing users to walk in the floor plan.
Milestone 1 (09.03)
- Research viability of different floor-plan parsing methods.
- Get started on backend. Project setup and a test frontend.
- Settle on the parsed file format.
Milestone 2 (23.03)
- Parse all types of image files as input. Convert them to PNG.
- Implement a standard CV algorithm for removing everything but walls from the input image. Should be able to adjust filtering based on the supplied image to work with all kinds of images.
Debug image illustrating floor plan parsing
Milestone 3 (06.04)
- Initial client setup:
- Image uploading from client.
- 3D walls reconstruction from parsed image.
3D viewer state
Milestone 4 (20.04)
- Basic editing capabilities in client:
- Delete walls/doors/windows.
- Place walls/doors/windows.
- Rotate walls.
- Scale walls/doors/windows.