Neural Earth Developer APIs

RoofAnalyticsAPI

Material class, structure type, footprint, and pitch-corrected surface area for any U.S. address. One endpoint. Sub-15-second response. Production-grade output your underwriting workflow can act on immediately.

One POST. Complete roof intelligence.

A single REST endpoint returns material class, structure type, footprint area, pitch-corrected surface area, capture date, and four debug images. Everything your workflow needs, nothing it doesn't.

Roof data you don't have to source, clean, or maintain.

Neural Earth handles satellite acquisition, segmentation, and classification. Your team gets production-grade output without building or maintaining any of the underlying data infrastructure.

Built for underwriting platforms at scale.

Designed for high-throughput ingestion into any PAS, rating engine, or underwriting workflow. If your team scores submissions programmatically, Roof Analytics drops in with a single integration.

Everything your underwriting workflow needs from a single address.

The Roof Analytics API takes a U.S. address and returns a complete GeoJSON Feature with material classification, structure type, footprint and pitch-corrected area, source metadata, and four Base64-encoded debug images per call — all in a single response, with a median latency under 15 seconds.

Roof material across 10 classes with calibrated confidence — Asphalt, Metal, Clay, Concrete, Membrane, BUR, Slate, Wood, Solar Panels, Solar Tiles
Roof structure across 9 classes — Gable, Hip, Cross-Hipped, Flat, Pyramid, Gable & Valley, Hip & Valley, Skillion, Saltbox
Footprint area (sqft/sqm) and pitch-corrected 3D surface area with correction factor
4 debug images per call — raw tile, masked tile, chipped region, raw chip — for underwriter review and audit
Capture date, geocode quality, zoom level, and model task manifest in every response

Three outputs that change how you score a roof.

Material Classification

Know the roof material before the policy is written.

A fine-tuned ResNet50 classifier returns the dominant roof material across 10 classes with a calibrated confidence score. Asphalt, metal, clay, concrete, membrane, BUR, slate, wood, and solar — classified from satellite imagery at the address level, every time.

Structure & Pitch-Corrected Area

Structure type and the surface area that actually matters for replacement cost.

The same call returns roof structure across 9 classes plus footprint area and pitch-corrected 3D surface area. SAM3 segments the building footprint; a structure-to-pitch heuristic converts it to actual roofing surface. Correction factor returned in every response.

Debug Imagery & Audit Trail

Every classification is reviewable and defensible.

Four Base64-encoded PNGs are returned per call — raw satellite tile, masked tile, chipped roof region, and raw chip — so underwriters can review the imagery behind any score and regulators can audit any decision.

Start scoring rooftops the same week you sign.

One endpoint. Production SLA. Into any underwriting platform. Talk to our team about volume, integration, and getting access.