OVER RESEARCH
Open Science for Spatial Intelligence and Physical AI
We publish our work in the open — papers, datasets and models built on the world's largest 3D map. Our research spans Large Geospatial Models, machine perception and spatial intelligence, bridging real-world 3D data and the AI systems that will act in the physical world.
Research Note
Sep 30, 2026
360° ReconstructionVisual-InertialMetric Scale
Can a 360° Camera Measure Its Own Scale?
Metric Scale for Insta360 Reconstructions from the Camera's Embedded IMU, Validated Against a Co-Mounted LiDAR
OVER Research
The 1 kHz IMU inside an Insta360 X5 gives our 360° reconstructions metric scale to 0.81% against a co-mounted LiDAR, 2.4× closer than the phone it replaced. On the way we found that the textbook visual-inertial alignment, solved over a whole survey, overestimates scale by 3.4% with a consistent sign; solving each contiguous run on its own and averaging removes the bias. On fifty production surveys (33 km) the public estimator ran unattended, passed its own checks every time and agreed with the phone to σ = 2.3%.
Research Note
Aug 28, 2026
Gaussian SplattingEmbodied AIVision-Language Models
A VLM Inside a 3D Capture Follows the Coordinate, Not the Room
Driving a Vision-Language Model Through a Gaussian-Splat Reconstruction, and Measuring How Much of It Is Vision
OVER Research
We put a robot pose inside a Gaussian-splat capture, rendered its view live at every step and let a vision-language model choose one action per frame. It works as a loop and produces a measured arrival — 19 steps, 1.03 m inside a 1.20 m radius, scored geometrically rather than self-reported — and a more useful negative result: a trivial compass follower reproduces 35–71% of the agent's actions, and if the goal's direction is withheld it never approaches the target, so this is coordinate-following with a visual veto rather than vision-and-language navigation. Two findings come with it: a harness that phrases privileged information as an instruction stops measuring the model, and a goal is only scorable if the capture walked inside its success radius (one of ours contained none of 438 capture poses). Pointed the other way, the same walk could hunt the poses a capture barely saw and hand them to a generative filler — stated at the end as a direction, deliberately not evidenced here.
Research Note
Jul 31, 2026
360° ReconstructionLIDARBenchmark
How Accurate Is Our Automatic 360°-Camera Reconstruction Pipeline?
Benchmarking the OVR 360° Pipeline Against a Simultaneous LIDAR Survey of the Same Site
OVER Research
We rode a LIDAR scanner and our 360° camera on the same pole through a 40 × 25 m site and compared the two reconstructions point by point: median surface agreement of 3.9 cm — at the reference scanner's own declared accuracy — drift-free geometry over the 118 m path, and one uniform 2.9% scale disagreement that the measurement chain traces to the metric reference rather than the reconstruction, removable with a single scalar.
Research Note
Jul 31, 2026
LGMMachine PerceptionFine-Tuning
Can 1,000 Open Scenes Make a State-of-the-Art 3D Model Better?
Assessing the OVER Dataset by Fine-Tuning a 3D Foundation Model on Its Open Sample
OVER Research
An empirical assessment of the OVER dataset for research, run entirely on OverMaps-1k — the open 1,000-scene sample anyone can download from Hugging Face: pose accuracy improves visibly on our domain, catastrophic forgetting is measured and defused through WiSE weight merging, and a scaling analysis shows how much data the gain requires.
Technical Report
DatasetLGMMachine Perception
The OVER 3D Maps Dataset
Internet-Scale Real-World 3D Data for Training Large Geospatial Models
OVER Research
A technical overview of the OVER 3D Maps Dataset: 270K+ real-world indoor and outdoor locations with 100M+ images, depth maps and metric scaling data — orders of magnitude larger than the datasets powering current Vision Foundation Models, and built for training Large Geospatial Models.
Open Dataset
DatasetOpen Source
OverMaps-1k
An Open Dataset of 1,000 Real-World 3D Maps
OVER Research
An open-source sample of the OVER 3D Maps Dataset released on Hugging Face: 1,000 real-world 3D maps with multi-view images, depth and metric scale, ready for benchmarking visual relocalization, depth estimation and 3D reconstruction models.
