Interactive Production Demo

From Street-Level Imagery to GIS-Ready Infrastructure Data

Geospatial Solutions extracts, classifies, and quality-assures right-of-way assets from street-level imagery — including traffic signs, streetlights, utility poles, fire hydrants, curb ramps, and more — then delivers validated, field-ready GIS datasets.

Detect & Annotate

AI-assisted detection with human annotation in CVAT

Validate & QA

Multi-pass quality assurance by dedicated reviewers

Deliver GIS-Ready

Enterprise-grade datasets in your preferred format

Interactive Asset Explorer

Explore extracted infrastructure assets across three Washington, D.C. corridors. Click any marker, bounding box, or table row to inspect an asset — selections synchronize across all views.

Geospatial View

Source Imagery

Select an asset on the map or table to view its source imagery

Confidence: 70%–100%

Asset ID	Type	Subtype	Corridor	Condition	Conf.	QC Status
GAV-BP-0001	Bike Lane Post	Tubular Marker	Georgia Avenue NE	Good	97%	Verified
GAV-BP-0002	Bike Lane Post	Tubular Marker	Georgia Avenue NE	Poor	93%	Verified
GAV-CB-0001	Catch Basin	Curb Inlet	Georgia Avenue NE	Poor	96%	Pending
GAV-CB-0002	Catch Basin	Combination Inlet	Georgia Avenue NE	Fair	95%	Verified
GAV-CR-0001	Curb Ramp	Depressed Corner	Georgia Avenue NE	Good	83%	Rejected
GAV-CR-0002	Curb Ramp	Blended Transition	Georgia Avenue NE	Good	92%	Verified
GAV-CR-0003	Curb Ramp	Parallel	Georgia Avenue NE	Fair	91%	Verified
GAV-CR-0004	Curb Ramp	Perpendicular	Georgia Avenue NE	Good	95%	Pending
GAV-CR-0005	Curb Ramp	Diagonal	Georgia Avenue NE	Fair	94%	Verified
GAV-CR-0006	Curb Ramp	Blended Transition	Georgia Avenue NE	Good	91%	Verified
GAV-FH-0001	Fire Hydrant	Dry Barrel	Georgia Avenue NE	Fair	88%	Verified
GAV-FH-0002	Fire Hydrant	Dry Barrel	Georgia Avenue NE	Poor	92%	Verified
GAV-FH-0003	Fire Hydrant	Dry Barrel	Georgia Avenue NE	Good	90%	Verified
GAV-FH-0004	Fire Hydrant	Flush-Mount	Georgia Avenue NE	Poor	99%	Corrected
GAV-PB-0001	Pedestrian Push Button	Standard	Georgia Avenue NE	Good	89%	Verified

Showing 1–15 of 155

1 / 11

Production Pipeline

A hybrid human-AI workflow where computer vision generates candidates and trained analysts produce the final, accountable data. Every record is reviewed before delivery.

Imagery Ingestion

Step 1 of 5

Street-level imagery from vehicle-mounted cameras or third-party providers (Cyclomedia, Nearmap) is ingested, georeferenced, and organized by corridor into the annotation platform.

Images tagged with GPS coordinates, vehicle heading, and capture timestamp
Organized into corridor-level batches for systematic, block-by-block coverage
Frame quality checks — blurry, overexposed, or corrupt frames flagged and excluded before annotation begins
Each frame indexed with a yellow-arrow vehicle marker and blue-cone facing direction for spatial context
Sequential frame ordering preserved so annotators can navigate forward/backward to find the best view of each asset
Metadata indexed for downstream traceability — every record links back to its source frame

Every frame is georeferenced and traceable from ingestion through delivery. No annotation begins until imagery passes quality checks.

QA/QC Process

Every record undergoes eight distinct quality checks before delivery. Flagged records enter a correction loop and are re-verified before inclusion in the final dataset.

Centroid Placement

Pass Rate: 96.8%

Verifies that the annotation centroid is placed at the visual center of the visible object footprint — not on the sign panel face, not on shadows, and not on pole tops. For long linear assets like guardrails and retaining walls, the centroid must be at the center of the visible segment. For sign posts, the point belongs at the post base footprint, not the sign face. Irregular shapes (Jersey barriers, catch basins) use the projected footprint center. Positional accuracy is validated within a 2-pixel tolerance. Calibration training ensures annotators apply consistent placement across all 14+ asset classes.

Error Report

7 records flagged during QA review

Asset ID	Type	Error	Severity	Description	Status
NCS-TS-0007	Traffic Sign	MUTCD Code Mismatch	Critical	Annotator assigned R2-1 (Speed Limit) but sign face shows W3-3 (Signal Ahead). Warning sign misclassified as regulatory — likely a copy-paste error from the previous annotation in sequence.	Flagged
GAV-SP-0003	Sign Post	Centroid Drift	Major	Centroid placed 4.2m from actual post location. Point was placed on the sign panel face rather than the post base footprint. Satellite imagery in this area obscured by tree canopy, requiring vehicle image as ground truth.	Flagged
RIA-CR-0002	Curb Ramp	Missing Attributes	Major	Detectable Warning Surface field left blank. Condition rated "Good" but source image shows cracked truncated dome panels with visible displacement. Ramp type field also defaulted instead of being explicitly set.	Flagged
NCS-FH-0002	Fire Hydrant	Duplicate Record	Minor	Same hydrant annotated in frames IMG-NCS-2024-0228-0018 and IMG-NCS-2024-0228-0019. Annotator did not select best image — labeled in both sequential frames instead of choosing the closest, clearest view.	Flagged
GAV-TR-0012	Street Tree	False Positive — Outside Zone	Major	Tree labeled behind residential fence, approximately 8m from road edge. Tree is not between sidewalk and road, not within 3m of road, and not in a median. Violates customer correctness rules for tree annotation zone.	Flagged
RIA-TS-0019	Traffic Sign	False Positive — Non-ROW Sign	Critical	Real estate "For Sale" sign annotated as traffic sign. Bounding box drawn around private property sign adjacent to ROW. MUTCD field set to "other" instead of being rejected as non-infrastructure.	Flagged
NCS-SL-0008	Street Light	Condition Rating Inconsistency	Major	Street light rated "Good" but source image shows visible base plate corrosion and the pole leaning approximately 15° from vertical. Distance blur may have caused annotator to overlook physical damage.	Flagged

Positional accuracy: 97.1% within 3mAttribute consistency: 98.2%False positive rate: 2.4%False negative rate: 4.7%Completeness rate: 95.3%Source-image traceability: 100%

Deliverables

A complete handoff package designed for immediate enterprise GIS ingestion. Every delivery includes spatial data, documentation, and a full audit trail.

File Geodatabase .gdb

Primary delivery format for ArcGIS environments. Contains feature classes with full attribute schema, spatial reference (NAD83 / EPSG:4269), and topology rules.

Esri File Geodatabase 10.x compatible
Point feature class per asset category
Attribute domains for condition, material, QC status
Spatial reference: NAD83 (EPSG:4269)
Metadata embedded per FGDC standard

Delivery Schema Preview17 fields

Field	Type	Description
asset_id	TEXT	Unique asset identifier (e.g., NCS-TS-0014)
feature_type	TEXT	Primary asset classification
subtype	TEXT	Sub-classification or MUTCD code
mutcd_code	TEXT	Manual on Uniform Traffic Control Devices code
material	TEXT	Construction material (Metal, Wood, Concrete, etc.)
condition	TEXT	Asset condition rating (Good / Fair / Poor)
shared_post	BOOLEAN	Whether asset shares mounting post
lat	FLOAT	WGS 84 latitude (decimal degrees)
lon	FLOAT	WGS 84 longitude (decimal degrees)
image_id	TEXT	Source capture frame identifier
capture_date	DATE	Date of street-level image capture
confidence	FLOAT	Detection confidence score (0.00–1.00)
annotator	TEXT	Annotator identifier (initials)
qa_reviewer	TEXT	QA reviewer identifier (initials)
qc_status	TEXT	QC review status (Verified / Corrected / Rejected / Pending)
qc_notes	TEXT	QA reviewer notes and correction details
delivery_date	DATE	Record delivery date

Proof of Value

One dataset delivery. Five operational outcomes. This data directly supports the programs and mandates your agency manages today.

Asset Inventory

A complete, attributed record of every inventoried right-of-way asset per corridor — traffic signs with MUTCD codes, street lights with pole types and mounting heights, utility poles with material classifications, fire hydrants, curb ramps, catch basins, and more. Establish a comprehensive baseline for capital planning, compliance, and resource allocation without sending field crews to every block. Each record includes condition rating, material, coordinates, source imagery, and a full QA audit trail.

Maintenance Planning

Standardized condition ratings (Good / Fair / Poor) with clear, repeatable definitions enable prioritized repair and replacement scheduling. Good = no visible damage. Fair = minor damage such as surface corrosion, moderate fading, or slight lean — still functional. Poor = major damage, broken, fallen, or unreadable. Focus maintenance budgets on the assets that need it most, backed by visual evidence, reviewer-verified assessments, and source-image traceability for every rating decision.

Corridor Analysis

Density maps, gap analysis, and MUTCD sign compliance reviews powered by spatially attributed data. Identify corridors with missing regulatory signage, insufficient street lighting coverage, aging infrastructure clusters, or sign posts with degraded reflectivity. Compare asset density and condition distributions across corridors to prioritize capital improvement zones. Every asset is geolocated and corridor-tagged for immediate spatial querying.

ADA / Accessibility Compliance

Curb ramp inventory with ramp type (perpendicular, parallel, diagonal), detectable warning surface status, and condition assessment supports ADA transition plan audits. Pedestrian push button inventory identifies locations with and without accessible pedestrian signals (APS). Flag intersections missing compliant ramps or lacking tactile indicators — directly from the delivered dataset, without additional field collection.

Emergency Readiness

Known asset locations, types, materials, and condition states support rapid damage assessment after storm events, flooding, or infrastructure failure. Pre-positioned data eliminates the need for emergency field surveys — dispatchers and assessment teams can query the GIS dataset immediately to identify affected assets by corridor, type, and condition. Fire hydrant locations, guardrail integrity, and street light status are available before crews reach the field.

This same data foundation can feed field mobile apps, public-facing dashboards, and resilience planning workflows — without any re-collection. One extraction. One QA process. Unlimited downstream applications.

Schedule a Consultation

This interactive demo represents the production workflow and data quality standards of Geospatial Solutions LLC. All data shown is synthetic and generated for demonstration purposes.

Explore More of Our Demos

This is just one of many interactive demos we've built. See our full collection of mapping, AI, and geospatial tools.

Browse All Demos

Technologies We Work With

Leveraging cutting-edge technologies and industry-leading tools to deliver exceptional geospatial solutions and data analytics services.

QGIS

GIS Software

ESRI ArcGIS

GIS Platform

PostgreSQL

Database

PostGIS

Spatial Database

AWS

Cloud Platform

Google Cloud

Cloud Platform

DuckDB

Analytics Database

OpenAI

AI Platform

Claude AI

AI Assistant

CVAT

Annotation Tool

Python

Programming

React

Frontend

Node.js

Backend

Docker

Containerization

Kubernetes

Orchestration

Azure

Cloud Platform

TensorFlow

Machine Learning

Pandas

Data Analysis

NumPy

Scientific Computing

Jupyter

Data Science

Git

Version Control

Linux

Operating System

Ubuntu

Operating System

Mapbox

Mapping Platform

Leaflet

Web Mapping

Fastapi

API Framework

GeoPandas

Geospatial Analysis

GDAL

Geospatial Library