Skip to content
Menu

AI solutions · Architecture AI

Floor-plan generation, spatial reasoning and CAD understanding

Public floor-plan sets are raster, research-only and not Indian. SVGZO's plans keep walls, openings and furniture as geometry and keep room labels, dimensions and schedules as text.

Fields that carry the supervision

  • rooms[] {label, area_sqft, bbox}
  • openings[]
  • dimension_system
  • unit_type
  • footprint
  • circulation

A training recipe

Build room-adjacency graphs from rooms[] and openings[] for graph-conditioned generation. Use unit_type and footprint as controllable conditions.

Evaluation

Buyers typically run DINO, SSIM, LPIPS and PSNR for image-to-SVG; FID, CLIP text-image score and render-validity rate for text-to-SVG; alignment and overlap metrics for layout; and IoU over room polygons for floor plans. A held-out evaluation set stratified by complexity is specified in the order, so these numbers stay comparable between runs.

Why vector data trains better models →