Read any radar into one model.
t3r, DZT, rd3, SEG-Y — raw survey files from any instrument are parsed into a single canonical representation. Add a new vendor and nothing downstream changes.
Xanadu · Substrata
Find the void before the road does.
Substrata reads ground-penetrating radar from any vendor into one model and surfaces subsurface cavities, pipes, and manholes — AI-assisted interpretation that runs in the browser, on the survey you already have.
Longitudinal, cross, and plan sections rendered live — AI detections drawn as you sweep.
Why Substrata
Built on one principle: what you train is what you infer.
01
Vendor-neutral
Raw survey files from any radar — read into a single canonical model. Add a new instrument, and detection, labeling, and export don't change.
02
AI that learns your ground
Pre-trained on physics-simulated radar, then sharpened on your own surveys. Every correction an analyst makes becomes training data.
03
Runs in the browser
No install, no vendor lock. Open a survey, sweep the sections, review detections in 2D and 3D — from any machine.
How it works
Every survey makes the next one sharper.
AI finds what the eye misses.
Pre-trained on physics-simulated radar, the model flags subsurface cavities, pipes, and manholes across longitudinal, cross, and plan sections — automatically, in seconds.
The analyst stays in command.
Every detection is a proposal, not a verdict. Accept a box with one click, correct it, or add what was missed — right there in the browser. No detection is ever trusted blindly.
Every correction trains the model.
Those clicks aren't discarded — they become labels. The model retrains on your ground, your instruments, your soil. Each survey makes the next detection sharper.
Capabilities
From raw trace to reviewed report.
Multi-view rendering
Longitudinal, cross, and plan sections with live A-scan and spectrum — the whole cube at a glance.
AI void & utility detection
Cavities, pipes, manholes and boxes flagged automatically. Accept a box with one click; it becomes a label.
2D → 3D fusion
Per-face detections fused into 3D boxes, positioned in real-world coordinates over the road.
On-screen calibration
Dewow, background removal, gain, zero-time — the same render pipeline used for training and inference.
Labeling built in
Draw, resize, reclass, undo. Annotations stored in canonical coordinates, upscale-invariant.
Export & interchange
Datasets and results out to standard formats — a bridge between instruments, not a silo.
Where it's used
Under every road that can't be allowed to fail.
- Road subsurface void & sinkhole precursor surveys
- Statutory underground-safety inspections
- Utility & buried-pipe mapping
- Manhole and buried-structure location
- Post-excavation verification
- Multi-vendor survey consolidation
See Substrata on your own survey.
Send us a few files. We'll run detection on your data and show you what it finds — no commitment.