Vestra
Scene contract, reconstruction, fusion, exports, and browser studio. The product owns the world and its provenance.
scene / studio ← Selected work A depth map
02 / SPATIAL COMPUTING CASE STUDY
A depth map
is not a world.
Vestra turns an ordinary phone video into an explorable spatial world — with the hard parts left visible: provenance, registration, residuals, and the limits of what the data can support.
LIVE SCENE CONTRACTLOCAL / INSPECTABLE
RGB VIDEODEPTH + POSEWORLD / EXPORT
01 / THE PROBLEM
Depth is a measurement.
A world is an argument.
A per-frame depth map can look convincing and still fail to describe one coherent room. Cameras drift. Scale changes between windows. A dense preview can hide broken registration. Vestra treats reconstruction as a chain of evidence: every derived surface stays attributable to the input and the transform that produced it.
02 / THREE REPOSITORIES
One product boundary, three explicit ownership boundaries.
Vestra Engine
Native Rust inference runtime for calibrated preprocessing, depth, confidence, and camera pose.
inference / CPUVestra Kernels
Shape-gated low-level CPU and experimental CUDA kernels, qualified independently before adoption.
kernels / parity03 / IMMUTABLE EVIDENCE
Make the world
auditable.
Raw measurements, derived geometry, and generated presentation are intentionally separate layers. A beautiful render cannot silently rewrite what was measured.
L0
Raw captureRGB frames / crop transform / device context
immutableL1
Depth + poseconfidence / intrinsics / camera residuals
traceableL2
Fused geometrysurfel field / TSDF / registration decision
derivedL3
Presentationprogressive reveal / browser studio / export
optional04 / MEASURED, NOT MARKETED
Parity first.
Then speed.
Same input, precision, and timed work within each slice. The figures below are separate benchmark scopes; they are not additive and do not describe an end-to-end product win.
RustC++ referenceRyzen 9 9950X · RTX 5080 · 96 GiB
Single-image CPU F32N = 20 · latency
28.3% lowerMultiviewN = 30 · complete slice
1.089% lowerGeometry + TSDFN = 10 · closed loop
4.11% lowerThe first slice also measured 39.5% higher throughput. These percentages stay scoped to their named protocol and should not be combined.
05 / RECONSTRUCTION IN MOTION
Not a render.
A recorded inspection.
Both clips come from the browser studio running the same frozen public scene. They are silent screen recordings, not edited geometry or a cinematic fly-through.
12 SEC / SILENT / 10 FPS
01 / MATCHED CAMERA
One timeline. Two inspectable views.
The original TUM RGB-D capture runs above Vestra's synchronized depth view. The displayed replay frames expose the crop, camera match, frame index, and 42,336 depth samples instead of hiding them behind a final render.
9 SEC / WEBGL VIEWER
02 / GLOBAL CONTROL
Measured points, freely inspected.
Vestra renders a 355,581-point geometric-MVS control derived by the pinned COLMAP provider and imported under the scene contract. The scene links 55 selected source frames; it uses no ground-truth camera poses and is not presented as pure Rust.
- Fixture
- TUM RGB-D / freiburg1_room
- License
- CC BY 4.0
- Local reconstruction
- 4,275,936 measured points / 9 windows
- Global control
- 355,581 fused MVS points / no GT poses
06 / HONEST LIMITS
What Vestra
does not claim.
01Relative scale is the current truth. A phone video does not become metric just because its point cloud is dense.
02There is no semantic mesh claim. Surfels and measured geometry are useful without inventing object labels.
03COLMAP is an optional, pinned provider for global registration experiments — not a pure-Rust implementation.
04A rejected pose hypothesis remains rejected evidence. Rendering more points cannot repair a broken global trajectory.
THE TAKEAWAY