26 Aug 2022 | Eric Kolve, Roozbeh Mottaghi, Winson Han, Eli VanderBilt, Luca Weihs, Alvaro Herrasti, Matt Deitke, Kiana Ehsani, Daniel Gordon, Yuke Zhu, Aniruddha Kembhavi, Abhinav Gupta, Ali Farhadi
AI2-THOR is an interactive 3D environment designed for visual AI research, available at <http://ai2thor.allenai.org>. It consists of near photo-realistic 3D indoor scenes where AI agents can navigate and interact with objects to perform tasks. Key features of AI2-THOR include:
1. **Interactions**: Supports various types of interactions such as object state changes, arm-based manipulation, and causal interactions.
2. **Scenes**: Provides a large number of interactive objects and scenes, including hand-modeled scenes by professional 3D artists and procedurally generated scenes.
3. **Quality**: The objects and scenes are near photo-realistic, facilitating better transfer of learned models to the real world.
4. **API**: Offers a Python API to interact with the Unity 3D game engine, enabling functionalities like navigation, object interaction, and physics modeling.
AI2-THOR is used for a wide range of tasks in Embodied AI, robotics, and computer vision, including visual navigation, learning object relationships, learning affordances, scene synthesis, and computer vision tasks. It has been used in over 150 publications and downloaded more than 500,000 times. The platform is highly customizable and supports various types of scenes, agent embodiments, actions, and metadata. AI2-THOR is continuously updated to add new features and improve performance.AI2-THOR is an interactive 3D environment designed for visual AI research, available at <http://ai2thor.allenai.org>. It consists of near photo-realistic 3D indoor scenes where AI agents can navigate and interact with objects to perform tasks. Key features of AI2-THOR include:
1. **Interactions**: Supports various types of interactions such as object state changes, arm-based manipulation, and causal interactions.
2. **Scenes**: Provides a large number of interactive objects and scenes, including hand-modeled scenes by professional 3D artists and procedurally generated scenes.
3. **Quality**: The objects and scenes are near photo-realistic, facilitating better transfer of learned models to the real world.
4. **API**: Offers a Python API to interact with the Unity 3D game engine, enabling functionalities like navigation, object interaction, and physics modeling.
AI2-THOR is used for a wide range of tasks in Embodied AI, robotics, and computer vision, including visual navigation, learning object relationships, learning affordances, scene synthesis, and computer vision tasks. It has been used in over 150 publications and downloaded more than 500,000 times. The platform is highly customizable and supports various types of scenes, agent embodiments, actions, and metadata. AI2-THOR is continuously updated to add new features and improve performance.