Home

What is the Spatial Intelligence Scene State Format?

• A lightweight, platform-agnostic, human and machine readable 3D scene representation standard
• A Python library for working with the Spatial Intelligence Scene State Format (pip install sissf)

Features:

• Parse and generate scene-state files with 3D model instances and transforms
• Support for architecture definitions (rooms, walls, floors, ceilings)
• Supports parent-child hierarchies for object relationships
• Compatible with multiple asset source (Objaverse, AI2-THOR, etc.)
• Geometry utilities for 3D bounding boxes, points, and transforms
• Quaternion-based rotations for accurate transformations

Components:

1. Scene State Format - specifies the set of 3D model assets used in the scene together with their transforms
2. Architecture Format - defines the building geometry of scenes, including rooms, walls, floors, ceilings, and openings (doors/windows)
3. Scene Graph Format - specifies the object-object and object-architecture relationships in the scene.

Preview

Scene State Format

{
  "format": "sceneState",
  "scene": { ... },      // Object instances and metadata
  "arch": { ... }        // Architecture
}

Python Library

from sissf import SceneState, Architecture, ModelInstance, ArchRoom

# Initialize scene
scene = SceneState(id="demo-scene")

# Create architecture
arch = Architecture(id="demo-arch")

# Create a room
room = ArchRoom(id="living_room")
arch.add_room(room)

# Add sofa to living room
sofa = ModelInstance(id="sofa-0", ...)
scene.add_instance(sofa)

# Export
scene.to_dict()

Credits:

• Inspired by the Smart Scenes Scene State Format
• Code implementation based on libsg