VisorVolt · Grid Inspection AI
THREE CAPABILITIES · ONE PLATFORM

Every asset you know. The ones you don't. And the defects.

See the assets — and defects — your inspections have been missing.

Capability 01

Standard assets

Standard grid hardware, detected directly.

raw frameRaw frame
Trained
detectorinstant
standard assets detected
Detected
Measured
81.0%
mAP@50
9 asset types · 422 held-out aerial frames
EPRI IEEE DataPort benchmark
Capability 02

Novel assets

Asset types no one labeled — learned live.

raw frameRaw frame
Auto-label flywheel
IDLE
 
novel asset detected
cutout 0.00
Discovered
Live capability
0
new labels
Novel types discovered & verified on aerial frames
93.8% verified precision
Capability 03

Defects

Component faults flagged, ranked, ready to action.

insulator frameRaw frame
Defect
classifierper-component
defect flagged
Insulator defect
Flagged
Measured
0.74
F1 · insulator defects
Broken disc · suspension string
Severity · HighPriority · P2
∞
Self-Improving Agent Autonomous LIVE IN DEPLOYMENT
Watches every result→Flags drift & errors→Retrains the model→Redeploys↻ continuous
A closed-loop agent reviews every model's output and retrains each one automatically — accuracy compounds with every flight, no engineer in the loop.
Accuracy compounds ↗
🧑‍🔧
Inspector
review
✓ Approve
Where it lands

Approved results flow straight into the systems you already run.

Esri ArcGIS · GIS SAP · assets IBM Maximo · work orders Compliance reports
Turn the drone imagery you already fly into a complete asset & defect inventory.
Defect F1 measured on CPLID benchmark (66% precision · 84% recall) · Novel-class 93.8% precision on capacitor bank (70% recall)