#55 · Primary category: Robotics & Embodied AI Simulation

Holodeck

3d-environment ai2-thor generative-ai large-language-models text-to-3d unity

CVPR 2024: Language Guided Generation of 3D Embodied AI Environments.

Project last updated:04/02/25

GitHub Stars

566

Forks

64

Contributors

5

License

Apache-2.0

Why we included this project

Holodeck takes a plain-language description of a room, like "a living room," and turns it into a furnished, navigable 3D scene for embodied AI agents. A large language model breaks the query into a list of objects and spatial constraints, and a constraint solver then places those objects into an AI2-THOR environment. That makes it handy for researchers who need varied test scenes for embodied agents or navigation work without hand-authoring each one. It runs locally on macOS or Ubuntu, and you can load the resulting scenes into Unity to inspect them. Expect a research-oriented setup: you'll install specific AI2-THOR versions and download the Objaverse-based asset data before generating your first scene.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category