Meet the Winners of Our First Scaniverse #SplatYourWorld Challenge | Niantic Spatial, Inc.

Every generation of technology has its breakthrough. Sprites to polygons. Handcrafted models to photogrammetry. Photogrammetry to neural fields. Now, the era of Gaussian splats.

In just a few years, splats went from research breakthrough to one of the most exciting advances in 3D reconstruction, powering photorealistic digital twins that change how we visualize, simulate, and interact with the physical world.

But no technological leap runs on technology alone. It runs on the community of early adopters and industry shapers who push the tools forward.

That's why we launched the first-ever Scaniverse #SplatYourWorld Challenge: to celebrate the creators driving Gaussian splats forward.

The response blew past our expectations. From natural landscapes and historic landmarks to complex industrial sites, the creativity and technical excellence were inspiring. Every submission is another step forward for 3D reconstruction, digital twins, and Spatial AI.

Overall Winner: The Benchmark for Gaussian Splats

Salar de Uyuni by Hiroto Otsubo

The Grand Prize recognizes the reconstruction that sets the benchmark for what's possible with Gaussian splatting today. Judges evaluated submissions for reconstruction quality, scale & coverage, and real-world impact

The “Salar de Uyuni” capture represents the world’s largest salt flat in southwest Bolivia and serves as a compelling example of how high-quality Gaussian splats can faithfully capture the complexity of the physical world while enabling the next generation of AI systems to perceive, navigate, and interact with real-world environments. As Judge Mohamed Sayed, a Senior Research Scientist at Niantic Spatial noted, "The overall winner is novel in a few ways: the scene itself is fascinating, an open expanse of almost nothing. In an area where people aim to cram the most detail per pixel, this scene shines with the opposite — gentle captivating textures.” Lighting and reflections can pose significant challenges for reality capture, but this entry turned those conditions into one of its greatest strengths.

Captured using an Insta360 X5 camera, Hiroto Otsubo reconstructed this unique and technically challenging natural landscape at a level that reflects the current state of the art in 3D reconstruction technology. In his own words, Hiroto “was inspired by the surreal scale of Salar de Uyuni, where an endless reflective surface stretches across the landscape more than 3,500 meters above sea level. With almost no buildings, trees, or distinct visual landmarks, it is also an unusually challenging environment for 3D Gaussian Splatting. I wanted to see whether Scaniverse could reconstruct not only the Land Cruiser, but also the vastness, reflections, and sense of isolation that make Uyuni feel unlike anywhere else on Earth.”

The result demonstrates what advanced Gaussian splatting can achieve even in environments where scale, reflections, and a lack of distinct visual features make accurate reconstruction especially challenging. Powered by Niantic Spatial’s reconstruction technology, which uses AI priors to help infer geometry and structure from visual data, captures like this push beyond traditional reconstruction techniques to create more complete, high-fidelity representations of the physical world.

Best Industrial Environment Reconstruction

Sleutel Heat Centre Leiden by Gabriele Romagnoli, XR AI Spotlight

Industrial environments are among the most demanding spaces to reconstruct. Dense machinery, repetitive structures, confined spaces, and complex geometry require a level of accuracy and completeness that extends beyond visual quality alone.

The aerial capture of the Sleutel Heat Centre Leiden, an industrial waste heating facility, was completed with an Antigravity A1 drone by Gabriele Romangnoli. “This reconstruction excels at thin, complex geometry and challenging materials: the top dome structures and scaffolding are rendered with impressive clarity, complemented by the subtle building reflections in the canal,” commented Judge Oliver Johnston, an AI Operations Specialist at Niantic Spatial.

Overall, Gabriele demonstrated how high-quality Gaussian splats can preserve the complexity of industrial environments while providing the spatial understanding needed for enterprise-scale inspection, automation, robotics, and AI-powered operations.

Best Natural Landscape Reconstruction

Detunata Goală by Razvan Onisca, Archaic Knowledge

Irregular terrain, dense vegetation, changing lighting, and organic geometry push Gaussian splats to their limits. The Natural Landscapes category recognizes submissions that faithfully construct digital representations that help preserve, monitor, and better understand the natural world.

Using a DJI Avata 360 drone, Razvan Onisca reconstructed Detunata Goală, a well-known geological site and nature reserve in Transylvania, Romania. Razvan shared, “I love how fast and easy it is working with Scaniverse, without sacrificing quality, and at very reasonable pricing.”

Judge Michael Rubloff, Founder of Radiance Fields, praised Razvan’s submission: “Detunata Goală is a perfect example of how lifelike 3D reconstruction can capture a place as it actually exists, bring people somewhere they'd never otherwise stand, and lay the groundwork for an entirely new imaging medium.”

Beyond visualization, 3D reconstructions of natural landscapes can support change detection and hazard awareness in remote areas where monitoring is difficult but mission-critical. Paired with localization technologies like visual positioning, they can also help people and machines navigate GPS-degraded or denied environments.

Best Urban Infrastructure Reconstruction

CORPUS Museum by Gabriele Romagnoli, XR AI Spotlight

Cities are where the future of Spatial AI comes to life. The Urban Infrastructure category recognizes exceptional reconstructions of streets, buildings, and public spaces that will help people, robots, and AI systems understand and navigate the real world.

The CORPUS Museum submission stood out for not only its exceptional reconstruction quality, but also its absolute scale and completeness featuring a striking 35-meter-tall seated human figure that dominates the urban skyline in South Holland. Judge Jonathan Stephens, Founder of Pixel Reconstruct stated, “I could see this used for a base of a 3D map,” highlighting how Gaussian splats can capture the richness and complexity of diverse environments, providing a foundation for world models, environmental understanding, and the next generation of Spatial AI.

Behind this reconstruction, Gabriele Romagnoli captured just 4 minutes and 25 seconds of 360-degree drone footage using the Antigravity A1. Processed in the cloud with Niantic Spatial’s Scaniverse, the footage was transformed into a high-fidelity reconstruction of this iconic cityscape in the Netherlands.

Looking Ahead

Every reconstruction contributes to richer world models that will enable people, robots, autonomous systems, and AI to better understand, navigate, and interact with the physical world. Together, we're building the real-world foundation models that will unlock Physical AI, creating a future where people and intelligent machines seamlessly interact with each other and the world around them.

Beyond our category winners, we’d also like to recognize two standout submissions with Honorable Mentions: Beatriz Junquera’s detailed reconstruction of the historic Bates-Scofield Homestead in Darien, Connecticut, and Justin Ryan’s large-scale aerial capture of downtown Chattanooga.

Thank you to everyone who participated. We can't wait to see what you create next!

About Niantic Spatial

Niantic Spatialは、フィジカルAIのためのリアルワールドモデルを構築しています。ロボット・AIエージェント・人々が物理空間を正確かつ共通に理解できるようにすることを目指しています。3D再構築技術はあらゆる標準的なカメラから幾何学的精度と精細なディテールで環境をキャプチャし、Visual Positioning Systemは世界中のほぼあらゆる場所で正確な位置推定を提供します。詳細はnianticspatial.comをご覧ください。