. Prior work has argued that machine learning may also benefit from interactive, multimodal learningHermann et al. , Oh et al. , de Vries et al. , termed virtual embodiment Kiela et al. . Driven by breakthroughs in static, unimodal tasks such as image classification Krizhevsky et al.  and language processing Mikolov et al. , machine learning has moved in this direction. Recent tasks such as visual question answering Antol et al. , image captioning Vinyals et al. , and audio-video classification Dhall et al.  make steps towards learning from multiple modalities but lack the dynamic, responsive signal from exploratory learning. Modern, challenging tasks incorporating interaction, such as Atari Bellemare et al.  and Go Silver et al. , push agents to learn complex strategies through trial-and-error but miss information-rich connections across vision, language, sounds, and actions. To remedy these shortcomings, subsequent work introduces tasks that are both multimodal and interactive, successfully training virtually embodied agents that, for example, ground language in actions and visual percepts in 3D worlds Hermann et al. , Oh et al. , Chaplot et al. .
For virtual embodiment to reach its full potential, though, agents should be immersed in a rich, lifelike context as humans are. Agents may then learn to ground concepts not only in various modalities but also in relationships to other concepts, i.e. that forks are often in kitchens, which are near living rooms, which contain sofas, etc. Humans learn by concept-to-concept association, as shown in child learning psychology Landau et al. , Smith and Yu , cognitive science Barsalou , neuroscience Nakazawa et al. , and linguistics Quine et al. . Even in machine learning, contextual information has given rise to effective word representations Mikolov et al. , improvements in recommendation systems Adomavicius and Tuzhilin , and increased reward quality in robotics Jaderberg et al. . Importantly, scale in data has proven key in algorithms learning from context Mikolov et al.  and in general Russakovsky et al. , Bojar et al. , Tobin et al. .
To this end, we present HoME: the Household Multimodal Environment (Figure 1). HoME is a large-scale platform111Available at https://home-platform.github.io/ for agents to navigate and interact within over 45,000 hand-designed houses from the SUNCG dataset Song et al. . Specifically, HoME provides:
3D visual renderings based on Panda3D.
3D acoustic renderings based on EVERT Laine et al. , using ray-tracing for high fidelity audio.
Semantic image segmentations and language descriptions of objects.
Physics simulation based on Bullet, handling collisions, gravity, agent-object interaction, and more.
A Python framework integrated with OpenAI Gym Brockman et al. .
HoME is a general platform extensible to many specific tasks, from reinforcement learning to language grounding to blind navigation, in a real-world context. HoME is also the first major interactive platform to support high fidelity audio, allowing researchers to better experiment across modalities and develop new tasks. While HoME is not the first platform to provide realistic context, we show in following sections that HoME provides a more large-scale and multimodal testbed than existing environments, making it more conducive to virtually embodied learning in many scenarios.
2 Related work
The AI community has built numerous platforms to drive algorithmic advances: the Arcade Learning Environment Bellemare et al. , OpenAI Universe Uni , Minecraft-based Malmo Johnson et al. , maze-based DeepMind Lab Beattie et al. , Doom-based ViZDoom Kempka et al. , AI2-THOR Zhu et al. , Matterport3D Simulator Anderson et al.  and House3D Anonymous . Several of these environments were created to be powerful 3D sandboxes for developing learning algorithms Johnson et al. , Beattie et al. , Kempka et al. , while HoME additionally aims to provide a unified platform for multimodal learning in a realistic context (Fig. 2). Table 1 compares these environments to HoME.
The most closely related environments to HoME are House3D, AI2-THOR, and Matterport3D Simulator, three other household environments. House3D is a concurrently developed environment also based on SUNCG, but House3D lacks sound, true physics simulation, and the capability to interact with objects — key aspects of multimodal, interactive learning. AI2-THOR and Matterport3D Simulator are environments focused specifically on visual navigation, using 32 and 90 photorealistic houses, respectively. HoME instead aims to provide an extensive number of houses (45,622) and easy integration with multiple modalities and new tasks.
Other 3D house datasets could also be turned into interactive platforms, but these datasets are not as large-scale as SUNCG, which consists of 45622 house layouts. These datasets include Stanford Scenes (1723 layouts) Fisher et al. [2012b], Matterport3D Chang et al.  (90 layouts), sceneNN (100 layouts) Hua et al. , SceneNet (57 layouts) Handa et al. , and SceneNet RGB-D (57 layouts) McCormac et al. . We used SUNCG, as scale and diversity in data have proven critical for machine learning algorithms to generalize Russakovsky et al. , Bojar et al.  and transfer, such as from simulation to real Tobin et al. . SUNCG’s simpler graphics also allow for faster rendering.
|Atari Bellemare et al. ||•|
|OpenAI Universe Uni ||•||•||•||•|
|Malmo Johnson et al. ||•||•||•||•|
|DeepMind Lab Beattie et al. ||•||•||•|
|VizDoom Kempka et al. ||•||•||•|
|AI2-THOR Zhu et al. ||•||•||•||•||•|
|Matterport3D Simulator Anderson et al. ||•||•||•||•|
|House3D Anonymous ||•||•||•||•||•|
Overviewed in Figure 1, HoME is an interactive extension of the SUNCG dataset Song et al. . SUNCG provides over 45,000 hand-designed house layouts containing over 750,000 hand-designed rooms and sometimes multiple floors. Within these rooms, of which there are 24 kinds, there are objects from among 84 categories and on average over 14 objects per room. As shown in Figure 3, HoME consists of several, distinct components built on SUNCG that can be utilized individually. The platform runs faster than real-time on a single-core CPU, enables GPU acceleration, and allows users to run multiple environment instances in parallel. These features facilitate faster algorithmic development and learning with more data. HoME provides an OpenAI Gym-compatible environment which loads agents into randomly selected houses and lets it explore via actions such as moving, looking, and interacting with objects (i.e. pick up, drop, push). HoME also enables multiple agents to be spawned at once. The following sections detail HoME’s core components.
3.1 Rendering engine
The rendering engine is implemented using Panda3D Goslin and Mine 
, an open-source 3D game engine which ships with complete Python bindings. For each SUNCG house, HoME renders RGB+Depth scenes based on house and object textures (wooden, metal, rubber, etc.), multi-source lighting, and shadows. The rendering engine enables tasks such as vision-based navigation, imitation learning, and planning.
This module provides: RGB image (with different shader presets), depth image.
3.2 Acoustic engine
The acoustic engine is implemented using EVERT222https://github.com/sbrodeur/evert, which handles real-time acoustic ray-tracing based on the house and object 3D geometry.
EVERT also supports multiple microphones and sound sources, distance-dependent sound attenuation, frequency-dependent material absorption and reflection (walls muffle sounds, metallic surfaces reflect acoustics, etc.), and air-absorption based on atmospheric conditions (temperature, pressure, humidity, etc.).
Sounds may be instantiated artificially or based on the environment (i.e. a TV with static noise or an agent’s surface-dependent footsteps).
This module provides: stereo sound frames for agents w.r.t. environmental sound sources.
3.3 Semantic engine
HoME provides a short text description for each object, as well as the following semantic information:
Color, calculated from object textures and discretized into 16 basic colors, ~130 intermediate colors, and ~950 detailed colors333Colors based on a large-scale survey by Randall Munroe, including relevant shades such as “macaroni and cheese” and “ugly pink,” https://blog.xkcd.com/2010/05/03/color-survey-results/.
Category, extracted from SUNCG object metadata. HoME provides both generic object categories (i.e. “air conditioner,” “mirror,” or “window”) as well as more detailed categories (i.e. “accordion,” “mortar and pestle,” or “xbox”).
Material, calculated to be the texture, out of 20 possible categories (“wood,” “textile,” etc.), covering the largest object surface area.
Size (“small,” “medium,” or “large”) calculated by comparing an object’s mesh volume to a histogram of other objects of the same category.
Location, based on ground-truth object coordinates from SUNCG.
With these semantics, HoME may be extended to generate language instructions, scene descriptions, or questions, as in Hermann et al. , Oh et al. , Chaplot et al. .
HoME can also provide agents dense, ground-truth, semantically-annotated images based on SUNCG’s 187 fine-grained categories (e.g. bathtub, wall, armchair).
This module provides: image segmentations, object semantic attributes and text descriptions.
3.4 Physics engine
The physics engine is implemented using the Bullet 3 engine444https://github.com/bulletphysics/bullet3.
For objects, HoME provides two rigid body representations: (a) fast minimal bounding box approximation and (b) exact mesh-based body.
Objects are subject to external forces such as gravity, based on volume-based weight approximations.
The physics engine also allows agents to interact with objects via picking, dropping, pushing, etc.
These features are useful for applications in robotics and language grounding, for instance.
This module provides: agent and object positions, velocities, physical interaction, collision.
Using these engines and/or external data collection, HoME can facilitate tasks such as:
Instruction Following: An agent is given a description of how to achieve a reward (i.e. “Go to the kitchen.” or “Find the red sofa.”).
Visual Question Answering: An agent must answer an environment-based question which might require exploration (i.e. “How many rooms have a wooden table?”).
Dialogue: An agent converses with an oracle with full scene knowledge to solve a difficult task.
Pied Piper: One agent must follow another specific agent, out of several, each making specific sounds. HoME’s advanced acoustics allow agents with multichannel microphones to perform sound source localization and disentanglement for such a task.
Multi-agent communication: Multiple agents communicate to solve a task and maximize a shared reward. For example, one agent might know reward locations to which it must guide other agents.
Our Household Multimodal Environment (HoME) provides a platform for agents to learn within a world of context: hand-designed houses, high fidelity sound, simulated physics, comprehensive semantic information, and object and multi-agent interaction. In this rich setting, many specific tasks may be designed relevant to robotics, reinforcement learning, language grounding, and audio-based learning. HoME’s scale may also facilitate better learning, generalization, and transfer. We hope the research community uses HoME as a stepping stone towards virtually embodied, general-purpose AI.
We are grateful for the collaborative research environment provided by MILA. We also acknowledge the following agencies for research funding and computing support: CIFAR, CHISTERA IGLU and CPER Nord-Pas de Calais/FEDER DATA Advanced data science and technologies 2015-2020, Calcul Québec, Compute Canada, and Google. We further thank NVIDIA for donating a DGX-1 and Tesla K40 used in this work. Lastly, we thank acronymcreator.net for the acronym HoME.
- Fisher et al. [2012a] Kelly Fisher, Kathy Hirsh-Pasek, and Roberta M Golinkoff. Fostering mathematical thinking through playful learning. Contemporary debates on child development and education, pages 81–92, 2012a.
- Landau et al.  Barbara Landau, Linda Smith, and Susan Jones. Object perception and object naming in early development. Trends in cognitive sciences, 2(1):19–24, 1998.
- Smith and Yu  Linda Smith and Chen Yu. Infants rapidly learn word-referent mappings via cross-situational statistics. Cognition, 106(3):1558–1568, 2008.
- Hermann et al.  Karl Moritz Hermann, Felix Hill, Simon Green, Fumin Wang, Ryan Faulkner, Hubert Soyer, David Szepesvari, Wojtek Czarnecki, Max Jaderberg, Denis Teplyashin, et al. Grounded language learning in a simulated 3d world. arXiv preprint arXiv:1706.06551, 2017.
- Oh et al.  Junhyuk Oh, Satinder Singh, Honglak Lee, and Pushmeet Kohli. Zero-shot task generalization with multi-task deep reinforcement learning. In ICML, 2017.
- de Vries et al.  Harm de Vries, Florian Strub, Sarath Chandar, Olivier Pietquin, Hugo Larochelle, and Aaron C. Courville. Guesswhat?! visual object discovery through multi-modal dialogue. In CVPR, 2017.
Kiela et al. 
Douwe Kiela, Luana Bulat, Anita L Vero, and Stephen Clark.
Virtual embodiment: A scalable long-term strategy for artificial intelligence research.In Machine intelligence workshop at NIPS, 2016.
- Krizhevsky et al.  Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. In NIPS, 2012.
- Mikolov et al.  Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeff Dean. Distributed representations of words and phrases and their compositionality. In NIPS, 2013.
- Antol et al.  Stanislaw Antol, Aishwarya Agrawal, Jiasen Lu, Margaret Mitchell, Dhruv Batra, C Lawrence Zitnick, and Devi Parikh. Vqa: Visual question answering. In CVPR, 2015.
- Vinyals et al.  Oriol Vinyals, Alexander Toshev, Samy Bengio, and Dumitru Erhan. Show and tell: Lessons learned from the 2015 mscoco image captioning challenge. TPAMI, 39(4):652–663, 2017.
- Dhall et al.  Abhinav Dhall, OV Ramana Murthy, Roland Goecke, Jyoti Joshi, and Tom Gedeon. Video and image based emotion recognition challenges in the wild: Emotiw 2015. In International Conference on Multimodal Interaction, 2015.
- Bellemare et al.  Marc G Bellemare, Yavar Naddaf, Joel Veness, and Michael Bowling. The arcade learning environment: An evaluation platform for general agents. JAIR, 47:253–279, 2013.
- Silver et al.  David Silver, Aja Huang, Chris J Maddison, Arthur Guez, Laurent Sifre, George Van Den Driessche, Julian Schrittwieser, Ioannis Antonoglou, Veda Panneershelvam, Marc Lanctot, et al. Mastering the game of go with deep neural networks and tree search. Nature, 529(7587):484–489, 2016.
- Chaplot et al.  Devendra Singh Chaplot, Kanthashree Mysore Sathyendra, Rama Kumar Pasumarthi, Dheeraj Rajagopal, and Ruslan Salakhutdinov. Gated-attention architectures for task-oriented language grounding. arXiv preprint arXiv:1706.07230, 2017.
- Barsalou  Lawrence W Barsalou. Grounded cognition. Annual Review of Psychology, 59:617–645, 2008.
- Nakazawa et al.  Kazu Nakazawa, Michael C Quirk, Raymond A Chitwood, Masahiko Watanabe, Mark F Yeckel, Linus D Sun, Akira Kato, Candice A Carr, Daniel Johnston, Matthew A Wilson, et al. Requirement for hippocampal ca3 nmda receptors in associative memory recall. Science, 297(5579):211–218, 2002.
- Quine et al.  Willard Van Orman Quine, Patricia S Churchland, and Dagfinn Føllesdal. Word and object. MIT press, 2013.
- Adomavicius and Tuzhilin  Gediminas Adomavicius and Alexander Tuzhilin. Context-aware recommender systems. In Recommender systems handbook, pages 217–253. 2011.
- Jaderberg et al.  Max Jaderberg, Volodymyr Mnih, Wojciech Marian Czarnecki, Tom Schaul, Joel Z Leibo, David Silver, and Koray Kavukcuoglu. Reinforcement learning with unsupervised auxiliary tasks. arXiv preprint arXiv:1611.05397, 2016.
- Russakovsky et al.  Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. ImageNet Large Scale Visual Recognition Challenge. IJCV, 2015.
- Bojar et al.  Ondřej Bojar, Rajen Chatterjee, Christian Federmann, Barry Haddow, Matthias Huck, Chris Hokamp, Philipp Koehn, Varvara Logacheva, Christof Monz, Matteo Negri, Matt Post, Carolina Scarton, Lucia Specia, and Marco Turchi. Findings of the 2015 workshop on statistical machine translation. In Workshop on Statistical Machine Translation, 2015.
- Tobin et al.  Joshua Tobin, Rachel Fong, Alex Ray, Jonas Schneider, Wojciech Zaremba, and Pieter Abbeel. Domain randomization for transferring deep neural networks from simulation to the real world. In IROS, 2017.
- Song et al.  Shuran Song, Fisher Yu, Andy Zeng, Angel X Chang, Manolis Savva, and Thomas Funkhouser. Semantic scene completion from a single depth image. CVPR, 2017.
- Laine et al.  Samuli Laine, Samuel Siltanen, Tapio Lokki, and Lauri Savioja. Accelerated beam tracing algorithm. Applied Acoustics, 70(1):172–181, 2009.
- Brockman et al.  Greg Brockman, Vicki Cheung, Ludwig Pettersson, Jonas Schneider, John Schulman, Jie Tang, and Wojciech Zaremba. Openai gym. arXiv preprint arXiv:1606.01540, 2016.
- Uni  Openai universe. https://universe.openai.com/, 2016.
- Johnson et al.  Matthew Johnson, Katja Hofmann, Tim Hutton, and David Bignell. The malmo platform for artificial intelligence experimentation. In IJCAI, 2016.
- Beattie et al.  Charles Beattie, Joel Z Leibo, Denis Teplyashin, Tom Ward, Marcus Wainwright, Heinrich Küttler, Andrew Lefrancq, Simon Green, Víctor Valdés, Amir Sadik, et al. Deepmind lab. arXiv preprint arXiv:1612.03801, 2016.
- Kempka et al.  Michał Kempka, Marek Wydmuch, Grzegorz Runc, Jakub Toczek, and Wojciech Jaśkowski. ViZDoom: A Doom-based AI research platform for visual reinforcement learning. In Computational Intelligence and Games, 2016.
- Zhu et al.  Yuke Zhu, Roozbeh Mottaghi, Eric Kolve, Joseph J Lim, Abhinav Gupta, Li Fei-Fei, and Ali Farhadi. Target-driven Visual Navigation in Indoor Scenes using Deep Reinforcement Learning. In ICRA, 2017.
- Anderson et al.  Peter Anderson, Qi Wu, Damien Teney, Jake Bruce, Mark Johnson, Niko Sünderhauf, Ian Reid, Stephen Gould, and Anton van den Hengel. Vision-and-language navigation: Interpreting visually-grounded navigation instructions in real environments. arXiv preprint arXiv:1711.07280, 2017.
- Anonymous  Anonymous. Building generalizable agents with a realistic and rich 3d environment. Under Submission at ICLR, 2018.
- Fisher et al. [2012b] Matthew Fisher, Daniel Ritchie, Manolis Savva, Thomas Funkhouser, and Pat Hanrahan. Example-based synthesis of 3d object arrangements. ACM Transactions on Graphics, 2012b.
- Chang et al.  Angel Chang, Angela Dai, Thomas Funkhouser, Maciej Halber, Matthias Niessner, Manolis Savva, Shuran Song, Andy Zeng, and Yinda Zhang. Matterport3d: Learning from rgb-d data in indoor environments. 3DV, 2017.
- Hua et al.  Binh-Son Hua, Quang-Hieu Pham, Duc Thanh Nguyen, Minh-Khoi Tran, Lap-Fai Yu, and Sai-Kit Yeung. Scenenn: A scene meshes dataset with annotations. In 3DV, pages 92–101. IEEE, 2016.
- Handa et al.  Ankur Handa, Viorica Pătrăucean, Simon Stent, and Roberto Cipolla. Scenenet: An annotated model generator for indoor scene understanding. In ICRA, 2016.
- McCormac et al.  John McCormac, Ankur Handa, Stefan Leutenegger, and Andrew J Davison. Scenenet rgb-d: Can 5m synthetic images beat generic imagenet pre-training on indoor segmentation? In CVPR, 2017.
- Goslin and Mine  Mike Goslin and Mark R Mine. The panda3d graphics engine. Computer, 37(10):112–114, 2004.