allenai/Holodeck
A CVPR 2024 research system that uses language models to procedurally generate 3D embodied AI environments for agent training and evaluation.

Velocity · 7d
+0.6
★ / day
Trend
→steady
star history
Holodeck enables language-guided generation of diverse 3D indoor environments for embodied AI research. Built on AI2-THOR and Objathor, it allows researchers to generate parameterized 3D scenes from textual descriptions, providing scalable training data for embodied agents. The system supports both macOS and Ubuntu and integrates with Unity for environment rendering.