HoME, the full name of which is Household Multimodal Environment, is a multimodal environment dataset designed to help artificial agents learn from vision, audio, semantics, physics, and interactions with objects or other agents in real environments.

HoME integrates SUNCG dataset45k+ different 3D house layouts, which can promote learning, generalization, and transfer-related learning.

HoME is an open source OpenAI Gym Compatible platform, extendable to tasks in reinforcement learning, language basics, sound-based navigation, robotics, multi-agent learning, and more.

The University of Sherbrooke released the system.