Aerial computer vision, at scale

Turn drone imagery into
actionable intelligence.

Heronflux is the inference platform built for aerial computer vision. Run your own models on gigapixel imagery — with geospatial primitives, multi-tenant isolation, and real cost attribution out of the box.

5–10×
Lower cost per inference
Gigapixel
Native imagery support
Multi-tenant
Isolation & attribution
The platform

Watch a model run on a live corridor.

Tile-aware scheduling, georeferenced outputs, GPU sharing — running on Heronflux right now.

Live app.heronflux.com / jobs / utility-corridor-NE-04
insulator_damaged 0.94
vegetation 0.87
corrosion 0.71
insulator_damaged 0.89
utility-corridor-NE-04
modelyolov8x-utility:v3.2
tiles144 / 196 · 73%
detections1,847
gpuA10 · healthy
cost$0.18 / 1k tiles
41.8024°N, 93.2104°W
EPSG:4326 · GSD 2.4 cm/px
Why Heronflux

One platform. One workload. Tuned end-to-end.

Horizontal infra can't optimize for what it doesn't know. We're built for aerial CV — and only aerial CV.

5–10× cheaper per inference

Tile-aware batching, weight caching across tenants, and GPU sharing with no quality loss. The economics that make per-customer inference viable.

Geospatial-native, not bolted on

Tiling, projection, and georeferencing are first-class. Inputs can be 50,000×50,000 pixels without you writing a single line of tiling code.

Multi-tenant isolation, real attribution

Per-tenant GPU quotas, weight isolation, and line-item cost attribution. Bill your end-customers what they actually consumed — to the cent.

Bring your own model

ONNX, TorchScript, or container. Pin a version per tenant. Cold-start, weight caching, and rollouts handled — you focus on the model.

How it works

Three steps from model to georeferenced output.

Drop into an existing pipeline, not replace it.

01 / UPLOAD

Bring your model

Push your YOLO weights as ONNX, TorchScript, or a container. Tag a version per tenant. Heronflux handles cold-start, weight caching, and rollouts.

02 / RUN

Point at imagery

COG, GeoTIFF, S3 URI, or a flight folder. Heronflux tiles with the right overlap, schedules across the fleet, and respects per-tenant quotas automatically.

03 / DELIVER

Real-world coordinates

Bounding boxes, masks, and classifications — all reprojected to the source CRS. GeoJSON, COG, or streamed to your pipeline via webhook.

Industries

Built for the workloads that move the world.

Energy, utilities, telecom, agriculture — anywhere aerial imagery meets computer vision.

Power transmission
Energy
Transmission & distribution
Solar farms
Renewables
Solar & wind asset audits
Agriculture
Agriculture
Crop health & yield mapping
Infrastructure
Industrial
Pipelines & towers
Cost attribution

Know exactly what each customer cost you.

Heronflux tracks GPU-time, tile-count, and inference latency per tenant — automatically. Bill your end-customers from real consumption data, not estimates.

Stop over-provisioning to cover the long tail. Start running tight margins on a fleet that does only what it needs to.

Aerial solar farm
Tenant utilization · last 30 days
acme-drone-services / production
End customer Tiles processed GPU-hours Utilization Cost
Midwest Power & Light 3.42M 184.7 94% $612.40
SolarCorp Northeast 1.18M 72.3 88% $238.12
Pipeline Integrity Co. 892K 54.1 91% $178.40
Telecom Tower Audit 421K 28.6 86% $94.20

Run your CV models the way they were meant to run.

See how Heronflux handles your imagery, your models, your tenants. We'll walk you through a live workload in 30 minutes.