FloorPlan Cleaner

A desktop preprocessing tool that converts IFC/BonsaiBIM-derived SVG floor plans—or floor plans drawn manually—into the Structured Scene Representation (SSR) consumed by the Opening-Aware, Style-Coherent Apartment Furnishing pipeline.

Features

• Detect enclosed rooms in IFC-annotated SVG floor plans.
• Assign room types and group rooms into apartment units.
• Preserve doors and windows in the exported SSR.
• Adjust detected room boundaries when the source drawing is ambiguous.
• Draw rooms and openings manually without an input SVG.
• Export room JSON files, unit metadata, and overview images for downstream furnishing.

Requirements

• Python 3.10 or newer
• A desktop environment capable of running PySide6

Quick start

git clone https://github.com/Thenolypus/FloorPlan-Cleaner.git
cd FloorPlan-Cleaner

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
python main.py

The activation command above is for Bash and Zsh. Fish users should run:

source .venv/bin/activate.fish

Clean an exported SVG

The SVG workflow is designed for 1:100 IFC/BonsaiBIM exports containing wall, door, and window metadata.

1. Select Open SVG and choose the exported floor plan.
2. Click inside each enclosed room to detect its boundary.
3. Select a room type and choose Assign Label.
4. Use Extend Boundary where a detected boundary needs correction.
5. After selecting all rooms belonging to an apartment, choose Save Unit.
6. Set the room height and choose Export All.

Repeat the room-selection and save steps before exporting when the SVG contains multiple apartment units.

Draw a floor plan manually

Select Create Unit to draw a floor plan without an SVG:

1. Draw and close each room polygon.
2. Add doors and windows to the appropriate walls.
3. Assign a room type to every room.
4. Adjust opening widths where necessary.
5. Choose Export Unit.

Manual projects can be saved and reopened. Exported metadata.json files can also be loaded back into Create mode for editing.

Export format

An SVG-based export produces:

<selected-output>/<floorplan-name>/
├── metadata.json
├── <floorplan-name>_centered.svg
└── unit_1/
    ├── unit_1_overview.png
    ├── unit_1_overview.svg
    ├── unit_1_room_1_<type>.json
    └── unit_1_room_2_<type>.json

Each room JSON contains its boundary, room type, architectural openings, and an initially empty object list. metadata.json records the room-to-unit mapping and the transforms required to reconstruct the complete apartment.

Pass the exported floor-plan directory to the furnishing pipeline:

uv run python -m src.run.pipeline \
  --floorplan-dir /path/to/<floorplan-name> \
  --unit 1 \
  --checkpoint /path/to/checkpoint \
  --arch

Roadmap

• Infer an editable initial sketch from PDF, JPEG, or PNG floor plans.

License

Released under the MIT License.