AI Quality Control & Defect QA — Snagging, Finishing & Handover

Challenges

Quality and handover live in the gap between the field and the office. Manual snagging is slow, inconsistent and easy to miss, and by the time defects surface the rework is expensive. Every minute a defect hides behind a finish or a void is a minute added to handover, and a defensible QA record is hard to build from paper punch-lists alone.

A · Manual snagging

Manual

A bare-headed worker without a hard hat among colleagues wearing helmets on a construction site

Manual snagging is slow

Late

A worker in plain clothes without a hi-vis vest standing near a moving wheel loader

Defects found after handover

Judgment

A worker at the unprotected edge of a high concrete slab without a fall-arrest harness

Inconsistent inspector judgment

B · Hidden defects

Hidden

Early-stage fire with small flames and smoke beginning among stacked site materials

Missed defects behind finishes

Sightline

An emergency exit doorway blocked by stacked boxes and debris

Blocked sightlines for inspection

Surface

An oil and water spill spreading across a concrete factory floor

Surface contamination & residues

C · Coverage gaps

Coverage

A forklift operating close to a pedestrian worker inside a warehouse

Can't inspect every unit

Concealed

An open excavation trench with no barrier or fence around it

Concealed service zones

Voids

A worker walking directly underneath a heavy load suspended from a tower crane

Ceilings & hidden voids

D · Documentation gaps

Records

A control room with one operator facing a large wall of many CCTV feeds

Paper punch-lists

Lighting

A large industrial site at night with only a few pools of light and large dark areas

Night & low-light defects

Lone

A single worker isolated in a remote corner of a large industrial site at dusk

Subjective quality ratings

E · Rework & handover

Curing

Outdoor workers under harsh midday desert sun with visible heat haze

Heat-cured defects

Reporting

A cluttered site safety office desk covered with paper incident forms and clipboards

No defect audit trail

Handover

A crowded site entrance at shift start with many workers entering at once

Slow handover & costly rework

Solutions

Every gap above has an answer below: how the AI sees, what it detects, how defect findings become action, the QA analytics you get, and a deployment that fits your existing cameras, site photos, 360° walk-throughs and Saudi data rules.

A · How the AI sees

Detection

Workers on a site each outlined by a translucent segmentation mask following their silhouette

Instance segmentation, not boxes

Precision

A worker with a glowing skeletal pose-estimation overlay analysing their posture

Crack- and surface-level precision

Defect

A worker with the helmet and vest highlighted by separate segmentation masks for PPE detection

Per-defect QA check

B · What it detects

A worker with helmet and vest highlighted by AI segmentation masks, unsafe-act detection

Concrete & Structure Defects

Flame and smoke highlighted by an AI segmentation mask, unsafe-condition detection

Finishing & Waterproofing Defects

Outdoor workers under harsh midday sun with a heat-stress overlay, poor-practice and compliance monitoring

Snagging & Handover Defects

C · From defect to action

Real-time

A site supervisor receiving a safety alert on a smartphone

Tiered, real-time defect alerts

Verified

Operators in a modern safety control room verifying alerts on a video wall

QA-verified

Dispatch

A public-address speaker on a pole at a site for live talk-down warnings

Route to the right crew

D · See it & prove it

Dashboard

Live QA operations dashboard

Analytics

A screen showing a site map with colour-coded risk heat zones

Defect trends & heat maps

Reporting

A tablet showing KPI gauges and charts for a safety compliance report

Defensible QA reporting

E · Fits your site

Integration An existing CCTV camera on a building with a small edge-AI box beside it

Works on existing CCTV & photos

Edge

Edge, hybrid or cloud

Residency

A secure on-premises server rack keeping data inside the Kingdom

KSA data residency

AI Technology

Our solution is built on data-driven video analytics, computer vision that watches live feeds and understands what it sees. The same AI runs across any input you bring: your existing CCTV, site photos, 360° walk-throughs, drones, autonomous robot dogs and smart helmets, all connected into one platform and one app, and processed on-site inside the Kingdom. We keep the engine proprietary; what follows is how it works for you.

Any input, one platform

AI Video Analytics

The brain behind every source, computer vision analyses each frame in real time and flags defects — from hairline cracks to spalling, paint runs and misaligned tile — the moment they appear.

Existing fixed and PTZ CCTV cameras on an industrial building

Your Existing CCTV

Runs on the cameras you already have, fixed, PTZ and thermal, with no rip-and-replace in most sites.

A survey drone flying over a construction site

Drones & Aerial

Drones carry the same AI across large sites and hard-to-reach areas, scanning facades, slabs and roof work for defects from above.

An autonomous quadruped robot dog patrolling an industrial site

Autonomous Robot Dog

A four-legged robot walks slab-to-slab on its own, taking defect scans where fixed cameras cannot reach.

A worker wearing a smart safety helmet with an on-board camera

Smart Inspection Helmet

Inspector-worn helmets with an on-board camera capture first-person QA scans of finishes, tile and ceilings, streamed to the same system.

On-Site Edge AI

All analysis runs on a secure unit at your site, so site photos, footage and defect data stay inside the Kingdom, in line with PDPL.

A unified safety app on a phone showing every connected camera and device

One Unified App, everything connected

Every camera, site photo batch, drone, robot and helmet feeds a single platform and one app, one place to watch the whole project, verify defect findings and act. Add a new input and it simply appears in the same view.

Integration with Your Systems

We adapt to your setup, you don't rebuild for us. The AI runs on the cameras and monitoring systems you already have: a small on-site unit pulls the video, your cameras and VMS keep running untouched, and in most sites nothing is replaced.

Works with what you have

A mix of existing fixed and PTZ CCTV cameras on an industrial building

No rip-and-replace

In 70–80% of sites your existing cameras are reused as they are. We connect over standard RTSP / ONVIF, compatible with virtually any IP camera.

A small edge-AI appliance plugged in beside an existing network switch

Plugs in alongside

A small on-site sidecar unit pulls the camera streams. Your cameras and current VMS keep running exactly as they do today, no disruption.

A technician surveying a CCTV camera with a tablet before deployment

Survey before spend

A pre-deployment camera survey (included) checks each camera's resolution, angle and lighting, so you know exactly what qualifies before buying anything.

Connects to your monitoring systems

Milestone XProtect Genetec Security Center Hikvision / HikCentral Dahua DSS Axis Camera Station

We integrate with the leading VMS platforms used in the Kingdom, pushing events and alerts straight into the system your control room already uses. Not on the list? The edge unit also runs fully standalone over direct RTSP, no VMS required.

Camera requirements

Parameter	Minimum	Recommended
Resolution	2 MP (1080p)	4 MP or higher
Frame rate	10 fps	15–25 fps
Low light	IR / night capable	True WDR
Detail on target	defect ≥ 5 mm	≥ 2 mm for hairline cracks
Compression	H.264	H.265 (½ the bandwidth)

Network & footprint

Light on bandwidth, a few Mbps per 1080p camera; H.265 halves it.
Only alerts leave the edge, short event clips and metadata, never full 24/7 video.
Processed on-site, a site VPN or private APN covers remote locations; data stays in the Kingdom.
No new software for your team, the dashboard is browser-based; nothing to install.

Use Cases

The same AI runs across every project — towers, villas, fit-out, precast and infrastructure. Every defect, from a hairline crack to a missed coat or a misaligned tile, is flagged on your site photos and cameras with a pixel-precise segmentation mask, never a vague note.

Concrete & Structure

AI segmentation mask on a surface crack in concrete — Surface Crack

AI segmentation mask on concrete spalling — Spalling

AI segmentation mask on concrete honeycombing — Honeycombing

AI segmentation mask on a concrete cold joint — Cold Joint

AI segmentation mask on exposed reinforcement bar — Exposed Rebar

AI segmentation mask on efflorescence salt deposits — Efflorescence

AI segmentation mask on concrete pop-outs — Pop-outs

Plaster & Paint

AI segmentation mask on a plaster crack — Plaster Crack

AI segmentation mask on debonded plaster — Plaster Debonding

AI segmentation mask on a paint run or sag — Paint Run / Sag

AI segmentation mask on paint blistering — Paint Blister

AI segmentation mask on an unpainted missed area — Missed Area

AI segmentation mask on a paint colour mismatch — Color Mismatch

Tiles & Stone

AI segmentation mask on tile lippage — Tile Lippage

AI segmentation mask on misaligned tiles — Tile Misalignment

AI segmentation mask on inconsistent grout joints — Grout Inconsistency

AI segmentation mask on a cracked tile — Tile Crack

AI segmentation mask on dirty grout — Dirty Grout

Waterproofing & Leaks

AI segmentation mask on a damp water stain — Damp / Water Stain

AI segmentation mask on an active leak — Active Leak

AI segmentation mask on a waterproof membrane gap — Membrane Gap

AI segmentation mask on failed sealant — Sealant Failure

AI segmentation mask on an incorrect floor slope — Floor Slope / Fall

Steel & MEP

AI segmentation mask on weld porosity — Weld Porosity

AI segmentation mask on weld undercut — Weld Undercut

AI segmentation mask on surface corrosion — Surface Corrosion

AI segmentation mask on misrouted electrical conduit — Conduit Misrouting

AI segmentation mask on a missing fixture — Missing Fixture

AI segmentation mask on an unsealed fire penetration — Penetration Not Firestopped

Snagging & Handover

AI segmentation mask on mechanical surface damage — Mechanical Damage

AI segmentation mask on incomplete work — Incomplete Work

AI segmentation mask on protruding fixings — Protruding Fixings

AI segmentation mask on construction debris — Construction Debris

AI segmentation mask on protective film left on — Protective Film Left

Not limited to these These are our standard library, we also train custom detection classes for the hazards, equipment and processes specific to your site.

How We Do It

01

Existing CCTV cameras on a pole overlooking an active construction site

Ingest

We pull live video from your existing CCTV, plus site photos and 360° walk-throughs, over RTSP, ONVIF or upload — no new cameras in most sites

02

A compact edge GPU appliance processing video on site

Edge AI

Inference runs on a small on-site GPU appliance, so defect scans, video and photos stay inside the Kingdom

03

A monitor showing workers segmented by AI masks on a live site feed

Detect

A three-stage cascade segments cracks, spalling, paint, tile, MEP and every other defect on every frame, across a broad class library

04

A control-room screen showing an AI reasoning check on a flagged event

Validate

A vision-language model reasons over context to confirm or suppress each defect finding, killing false calls

05

Alert

Tiered defect alerts reach the QA team, supervisors' phones and WhatsApp within seconds

06

Confirm

For non-critical findings a QA operator confirms with a clip and scene description before the repair crew is dispatched

07

Dashboard

Every camera, photo batch, unit, zone and live defect finding appears on one dashboard, on a wall screen or a phone

08

An automatically generated safety report with charts and an audit trail

Report & audit

Shift and daily defect reports, QA KPIs and an immutable audit trail are generated automatically for ISO 9001 and PDPL

The QA Platform

One console for the QA team. Overview, defect findings, cameras, units, zones, reports and live AI inspection, all in a single app. Every defect traces back to the photo or camera that raised it, the unit and zone it sits in, the moment it happened, and the confidence the AI gave it. Live today at viewkeepersa.com/VisualDefectDemo/qa.

QA app overview: quality score, defect breakdown, hottest zones and 3D zone map

QA Monitoring · Overview

Live AI defect detection: cracks, spalling, paint, tile, MEP, snagging & more
your PTZ, fixed and thermal cameras, plus uploaded site photos and 360° walk-throughs, on one live wall
Per-camera health, per-unit QA status and link status
QA verify, dispatch repair crew, and false-call workflow
Owner-ready daily, handover, and defect-evidence reports
Runs in the on-site control room, no foreign cloud

Platform Walkthrough — Page by Page

07 · A

Seven pages, live in our QA app today. The images below come straight from the running build. These are the same screens your quality team will open on day one.

viewkeepersa.com/VisualDefectDemo/qa

01 · Overview

The whole project, one screen

The QA landing page. A live quality score, defect compliance, open snags and critical findings run across the top, next to cameras online and mean response time. Below sits a 24-hour defect breakdown, the hottest zones ranked, and a 3D quality-zones map of the site.

Quality ScoreDefect BreakdownHottest Zones3D Zones Map

viewkeepersa.com/VisualDefectDemo/qa/events

02 · Findings

Every defect finding, triaged

The working queue of QA findings. Each one tagged by defect type, with the photo or camera and unit & zone that caught it, a severity, a confidence score and a status. An inspector verifies, assigns or closes, and the response clock keeps the team honest.

Defect ClassesSeverity RoutingEvidence SnapshotResponse Clock

viewkeepersa.com/VisualDefectDemo/qa/cameras

03 · Cameras

The fleet, at a glance

Every connected camera by site and zone, with live thumbnails, online status and the defects it raised. A technician sees what needs attention before the morning round.

Live ThumbnailsPer-ZoneOnline StatusLast Defect

viewkeepersa.com/VisualDefectDemo/qa/units

04 · Units

Per-unit QA, watched

Per-unit quality built for handover: defect density per villa or apartment, finishing-pass status, MEP readiness, and snag list completeness, so the handover record is built automatically and is always audit-ready.

Defect DensityFinishing PassMEP ReadinessSnag List

viewkeepersa.com/VisualDefectDemo/qa/zones

05 · Zones

Quality zones, drawn and enforced

Quality zones drawn on the site plan and the 3D model: per-floor, per-trade and per-finishing-package areas. Each zone carries its own QA rules, and any defect shows up in Findings with the zone attached.

Zone Rules3D ModelPer-TradeDefect Link

viewkeepersa.com/VisualDefectDemo/qa/reports

06 · Reports

Ready for the audit

Shift and daily summaries, defect exports and snag evidence packs, generated automatically and exportable for ISO 9001-aligned quality reviews and handover files. One click, and the record is in your inbox.

Shift SummaryDefect ExportEvidence PackISO 9001

viewkeepersa.com/VisualDefectDemo/qa/inspect

07 · Inspect

Point the AI at anything

A live AI scan page: drop in site photos or pick a camera, and the engine outlines cracks, spalling, paint, tile, MEP and every other defect on the spot. The fastest way to see the detection quality on your own scenes.

Live ScanAny FootageInstant MasksYour Scenes