SceneBloom
What it is
SceneBloom is an agentic framework that generates simulation-ready 3D indoor environments for embodied learning. It uses a two-stage pipeline: first, it samples visual layouts using FLUX and a vision-language model (Qwen3-VL) to define functional zones, and second, it compiles these proposals into a 3D scene using contract-guided lifting, fitting, and refinement steps.
The problem it solves
The lack of realistic and diverse 3D indoor environments for training and evaluating embodied AI agents.
What you could use it for
- Generating diverse training environments for embodied AI agents
- Creating simulation-ready 3D scenes for Habitat-sim
- Procedural generation of realistic room layouts (bedrooms, offices, etc.)
Why it matters
Provides a scalable method to create realistic, diverse, and simulation-ready indoor environments, addressing the data scarcity problem in embodied AI research.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.