Announcing the Open-Source Caldera Map Release

With a vision to expand the knowledge base of the gaming industry, Activision is excited to share the release of an open-source data set featuring the map Caldera from Call of Duty®: Warzone™ for non-commercial use.

This first-of-its-kind data set release for Call of Duty, now available in OpenUSD, contains the near-complete geometry of Caldera as well as a collection of randomly selected anonymized time samples showing how players move around the map. This initiative represents Activision’s commitment to innovation and growth in multiple fields, including AI learning, within the communities of academia and research, while propelling the advancement of game development. It brings to life the colossal and carefully crafted map that was home to many epic brawls in Call of Duty: Warzone and is production-proven and battle-tested.

Why Open-Source Matters

In an era where AI training and the evolution of authoring tools are pivotal, the availability of production-proven maps is crucial. By releasing Caldera as an open-source asset for non-commercial use, the technology teams at Activision aim to empower developers and educators with high-quality, production-validated and accessible resources. This is about collaborating with the gaming and research community to build a foundation for responsible innovation and learning across the industry.

A Resource for Education and Training

Our commitment to open source goes beyond development; Activision envisions these assets as invaluable educational resources. Students, educators, and professionals can utilize Caldera for training, demonstrations, and experimentation. This initiative encourages learning and skill development within the gaming community and beyond.

Supporting AI Development

Open-source assets like Caldera play a vital role in the advancement of artificial intelligence. By providing a rich, diverse environment, we facilitate the training of AI models, enhancing the industry’s understanding of complex geometries and interactions. This can lead to more intelligent systems, paving the way for the next generation of gaming and simulation technologies.

Evolving Authoring Tools

The gaming industry is consistently evolving with new authoring tools emerging regularly. By contributing assets to the open-source community, the Activision technology teams encourage the development of robust tools that can handle complex geometries and environments. This collaborative approach will drive innovation, making it easier for developers to create immersive experiences.

In a brief discussion with Activision’s Chief Technology Officer Natalya Tatarchuk and Senior Vice President, Fellow Software Engineer Michael Vance, the technology team details everything included in the set, the vision for the release, and how providing Caldera to academics and researchers can benefit the Call of Duty Community and positively impact the future of gaming.

What is the size and scale of the Caldera Data Set?

Michael Vance: The Call of Duty Data Set featuring Caldera represents an extensive, in terms of world-size, scene-graph depth and geometric complexity, production-quality map, used for multiplayer games in Call of Duty: Warzone. The release of this map represents one of the largest production-validated open data set releases from the gaming industry in terms of complexity of geometry and instance counts. It also is one of the largest publicly available OpenUSD data sets, providing an excellent test harness for OpenUSD itself.

Natalya Tatarchuk: We, at Activision, believe that it’s important for the gaming industry to help foster growth and innovation within the industry, to help evolve authoring tools, as well as provide excellent data for AI training and evolution of content generation techniques, and we believe that this data set provides a unique benefit for these goals.

The map's geometry data is approximately 4 GB but is comprised of more than five million meshes, 28 million primitives, and more than one-billion-point instances, which can also represent scene metadata such as volumes that we use for lighting processing. Its complexity is also a testament to the richness and detail that the team at Raven Software built, with help from Beenox, High Moon Studios, and the broader Call of Duty development teams.

Besides the geometry representing the Caldera map, what other data did you include in this data set?

Vance: We have also included in-game character pathing and time samples showing how players behave on the map. For instance, one of the sets shows the paths players take throughout a match. While we have not included specific visualizations of these, the data is easily accessible, allowing different ways to explore and visualize them.

Why was Caldera selected as the map to release as part of this set?

Vance: As our team looked through a variety of scenarios for the best data set, including multiple smaller multi-player maps, our priority was to choose a map that best represents the scale and complexity of our current design philosophy. It’s exciting for us to see Caldera continue to live on and help expand the games industry through academia and research. The hierarchical arrangement of the map also allows users to select sub-sections of the map to limit the scope of analysis or providing multiple smaller areas for comparison, which is useful in avoiding over-fitting issues by leaving other sections of the map for validation.

What do academics and researchers typically do with data sets such as this?

Tatarchuk: Researchers are constantly in search of data sets that represent the fullness and complexity of modern-day production data, which is difficult for them to approximate in a traditional academic setting.

The computer graphics industry has a strong legacy of canonical data sets—some may be familiar with for example the Stanford bunny, Dabrovic Sponza, or more recently Disney’s Moana Island, or Animal Logic’s ALab. Each of these has become an important part of research in the broader graphics and gaming industries, fostering innovation in areas such as mesh simplification, deferred rendering, and other modern lighting and shading approaches. With the lack of available production data at this scale, Caldera provides researchers an opportunity to explore these types of sets with more rigor.

What is your vision for what can come from the Caldera Data Set being released publicly?

Tatarchuk: Activision is committed to continuously being innovative in technology and AI space to further the development of video games and to help drive the gaming community forward.

Responsible innovation in AI technologies can help propel industry-wide development of video games forward in a way that creates richer gaming experiences for players in the future.

One element of releasing Caldera is giving back to the community of research that has been so beneficial to the industry over the past decades. One of our goals is to broaden access to production data, allowing researchers to test their approaches in real-world scenarios, which will help accelerate the development of new solutions, as well as provide confidence that solutions can be deployed at production scale. We believe that Caldera’s release will be an impactful and material benefit to that effort.

Vance: We are also interested in what new insights the research community will develop around object relationships, scene creation and editing, and other topics related to environmental construction with Caldera. The tabular data and metadata include providing opportunities for richer understanding of how players interact with the game, which can potentially lead to new insights on what makes specific layouts compelling for users.

As we learn what the research and academic community finds useful in the set, we want to create an ongoing dialog with them so we can continuously make updates that are helpful.

How can new technology or learnings from the Caldera Data Set benefit the Call of Duty community?

Vance: While our game environments are already enormously complex, we feel a constant need to improve the play experience and deliver even more richness and detail. Innovations that come from this data set release could give more freedom and flexibility for our content teams to find the most engaging scenarios for our players. Insights into object relationships, procedural approaches to our world data, and other ideas could lead to more compact data representations on both disk and in memory. We are excited about what the academic community will produce based on Caldera and look forward to seeing new techniques that we could never have imagined.

Join the Movement

We believe that the industry can greatly benefit from more open-source contributions. Our motivation stems from a shared desire to inspire creativity and collaboration. We invite everyone—students, studios, vendors, partners, and beyond—to explore, utilize, and contribute to the further growth of our incredible industry.

Download the Call of Duty Data Set Now

Join us in this exciting journey and become part of a vibrant community dedicated to innovation, education, and collaboration.

Together, let’s shape the future of gaming and beyond!

© 2024 Activision Publishing, Inc. ACTIVISION, and CALL OF DUTY are trademarks of Activision Publishing, Inc. All other trademarks and trade names are the property of their respective owner