Scene to layout

Skill wjlgatech/design-anything/skills/scene-to-layout

Turn a room photo, scan, or point cloud into a structured layout — walls, openings, furniture as typed objects with dimensions — using SpatialLM-class models, ready for blueprint-validate and re-design. Trigger: "here's a photo of my room", "measure this space", "redesign from this picture".From its SKILL.md

Install

npx -y skills add wjlgatech/design-anything --skill scene-to-layout

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its file declares

Copied from the file, not written here

The file declares its own license as CC-BY-4.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

1.8 KB, 344 tokens by cl100k_base, as published. Nobody here has run it

scene-to-layout

When to use

The input is a picture (the repo's second input modality) rather than text — an existing room/site that must become structured data before redesign.

What it does

Route by input: photo → monocular layout estimation; scan/point-cloud → SpatialLM (https://github.com/manycore-research/SpatialLM); floor-plan image → wall/opening extraction.
Emit a structured layout JSON: walls (start/end/height/thickness), openings (type/position/size), furniture (class/bbox/orientation), declared units.
State per-element confidence; dimensions estimated from a single photo are marked ESTIMATED and need one user-confirmed reference measurement.
Hand off to blueprint-validate (audit) or brief-to-blueprint (redesign).

Example

"Photo of a 4x5m living room + 'the door is 80cm' reference" → layout JSON with scaled dimensions → "now add a reading corner with daylight" → redesign loop.

Verification (eval-with-teeth)

Output JSON must round-trip: rendered back to a 2D plan, wall topology is closed (rooms are bounded polygons) and openings lie on walls. An unclosed room polygon is a failed extraction — report it, don't patch it silently.

Safety

Never claim measured accuracy from an unreferenced photo; renovation decisions need a tape measure.

Cross-runtime

Model access varies by runtime; degrade to "ask user for measurements" when no vision/SpatialLM backend is available (skip-not-fail).

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.