StreetProof™
Claim

StreetProof Physical Brand Index™ (PBI)

The technology behind physical citations

Evidence-firstPrivacy-scrubbedMachine-readable

StreetProof is a live evidence system, not an image-labeling demo. It turns eligible observations of storefronts, wrapped vehicles, signs, billboards, and other visible commercial assets into governed physical citations. Those citations build longitudinal Presence Records and the StreetProof Physical Brand Index™ for human and machine retrieval.

The system combines visual AI with geospatial context, entity and asset resolution, privacy controls, provenance, review, and an append-only ledger. The model output is never the final authority. Uncertain candidates remain private until the evidence gates required for public release are satisfied.

LIVE
public beta
REST + MCP
machine delivery
HASH-CHAINED
evidence ledger

The production pipeline

Each stage narrows what can move forward. Quality gates reduce noise, privacy processing protects people, entity resolution establishes what the scene supports, and evidence governance determines whether a candidate can enter the public record.

Capture and select

The Scout app and approved collection sources submit eligible public-space imagery. Quality and scene-change gates remove blank, unusable, and repeated frames before expensive processing.

Protect privacy

Faces, licence plates, and other sensitive details pass through the privacy pipeline before an image can become retained evidence or appear on a public record.

Read the scene

Computer vision, OCR, visual retrieval, and deterministic parsers identify commercial assets and extract candidate names, domains, phone numbers, and visible text.

Resolve the entity

Multiple identity signals connect an observation to the correct organization and asset. Models propose candidates; governed evidence rules and recorded review decide what can publish.

Remove duplicates

Image and asset matching stop repeated frames from becoming artificial evidence. Longitudinal observations remain distinct only when they document a valid event over time.

Publish and serve

Accepted evidence becomes a physical citation in the append-only ledger, then reaches people and machines through Presence Records, structured data, REST, MCP, and change feeds.

Privacy before publication

Faces and licence plates are blurred before any image becomes retained evidence, enters the recognition archive, or appears publicly. Unsanitized frames are processed transiently in the private privacy pipeline and are not retained as evidence.

Public records describe commercial evidence at an appropriate geographic scope. They do not publish private routes, precise source locations, raw capture files, faces, licence plates, or private Scout identifiers. If a required privacy or evidence gate fails, publication fails closed.

One ledger, several products

Physical citations

Each accepted event records what was observed, when it was observed, the supported presence kind, provenance, and the limits of the claim.

Presence Records

Entity pages assemble accepted citations into a longitudinal record without turning a vehicle, billboard, or sponsorship into a false address.

PBI and PPS

Eligible records receive a cohort-relative Physical Presence Score. Payment, claiming, and membership cannot create evidence or purchase rank.

Machine interfaces

Public HTML, structured data, REST, MCP, OpenAPI, and feeds expose the evidence with freshness, coverage, provenance, and claim boundaries attached.

Self-hosted NVIDIA accelerated computing

StreetProof runs central visual-AI, recognition, embedding, and entity-resolution workloads on controlled, self-hosted infrastructure powered by NVIDIA RTX PRO 6000 Blackwell. This supports local model development, inference, archive reprocessing, and multi-stage vision workloads while keeping sensitive processing under StreetProof's operational control.

In production now

Image-quality selection, privacy screening, commercial-asset detection, scene-text recognition, visual retrieval, entity resolution, duplicate detection, confidence scoring, and review routing.

NVIDIA roadmap

StreetProof is evaluating TensorRT for inference optimization, Triton for governed model serving, DeepStream for future multi-stream ingestion, and Jetson for privacy and candidate filtering at future fleet nodes.

Roadmap technologies are not presented as deployed production systems. The immediate engineering goal is higher throughput, lower processing cost per accepted citation, stronger resilience, and measured scaling without weakening privacy or publication controls.

The moat is the record

Models can be replaced. Observation history cannot be recreated after the fact. StreetProof compounds value through accepted citations, repeat observations, distinct assets, entity and brand relationships, reviewed recognition decisions, geographic coverage, provenance, and machine distribution. Together they form a physical-world data asset that web data alone cannot reproduce.

Inspect the public system

Read the methodology, inspect the ledger checkpoints, evaluate search and AI readiness, or work directly from the public API documentation.