INDUSTRIAL INTELLIGENCE

Industrial Vision & Edge AI

Negency brings AI to the factory floor and the edge: starting from scenario validation and technical design, controlled pilots test whether an industrial AI or edge deployment works — verified results stated plainly, unverified scope never promised.

Assess an industrial AI scenario
Industrial Vision & Edge AI
Photo: inspection camera inside a solder-paste inspection machine, Nelatan · Wikimedia Commons, CC BY-SA 3.0 (illustrative photo, not Negency equipment)

Industrial vision: verify first, then deploy

Visual inspection on a factory floor is not a demo: for every image judged defective, it must be possible to replay why it was judged so, who reviewed it, and which work order it entered. Black-box scores do not belong on a factory floor. Negency's approach is to state the method first, then prove it with a controlled pilot — pilot data decides whether the scheme continues.

The current status, stated plainly: this page is a technical scheme and a pilot method, not delivered capability. The first offline validation run is complete (2026-11-13, DeepPCB public benchmark: AUROC 0.627, detection rate 11.8%, false-positive rate 6.6% — weak numbers, a measured floor, not a capability claim); no real-device testing has taken place, and no edge-device pilot or site acceptance has begun. The verification record is published in the industrial vision sample in the Engineering Lab — what has been measured, and what is still missing, is written there.

The six-step loop

Image capture (industrial cameras and sensors) → edge inference (models running locally on edge compute modules) → defect analysis (defect-type judgement + confidence + evidence image) → evidence archiving (original image + inference result + review record, kept together) → human review (a mandatory step — the machine never clears a defect alone) → work order (confirmed defects enter the work-order and maintenance system). Deployment shape follows the client site: public cloud, client private cloud, on-premises, or edge devices can all be assessed; data can stay inside the plant. The reference architecture is diagrammed below.

Offline validation first

Before a model reaches an edge device, it is validated offline on the test set: detection accuracy, false-positive rate, and miss rate are measured, with thresholds set by acceptance metrics, not by demo effect. Datasets are licensed public datasets or client-authorised data only; defect categories and labelling rules are fixed before model selection; train, validation, and test sets are separated, and the test set plays no part in tuning before acceptance. Only after offline validation passes does the model move to an edge-device pilot.

Pilot path: four gates

Dataset selection → offline validation → edge-device pilot → site acceptance. Each step gates the next; pilot results decide whether site acceptance proceeds. Acceptance metrics (detection accuracy, false-positive rate, miss rate, per-image inference latency, continuous-run stability) and their thresholds are agreed with the client before the pilot starts and go into the acceptance document; a measurement that misses reworks the scheme instead of lowering the bar. The pilot plan and acceptance gates are tabulated below.

Capability boundaries

No promises on detection accuracy, false-positive rate, or deployment schedule — those await pilot measurement. An inspection verdict never triggers equipment action directly; no defect verdict reaches a work order without human confirmation; safety-related judgements are never left to the model alone. Equipment certification, functional safety, cybersecurity, and import/export compliance must be checked per product; no device is promoted as "fully compliant with EU and US regulations".

FIG. VSix-Step Inspection LoopNegency — Industrial Vision / Architecture
01

Image capture

Capture

Industrial cameras and sensors

02

Edge inference

Inference

Models run locally on edge modules

03

Defect analysis

Analysis

Type + confidence + evidence image

04

Evidence archiving

Evidence

Original + result + review record

05

Human review

Approval Gate

The machine never clears a defect alone

06

Work order

Work Order

Confirmed defects enter work orders

Steps and gate taken from the body above; rendered in code, no illustration. Missed samples feed back into the labelling rules — see Failure modes below.

METHODDecisions, selection, pilot & acceptanceNegency — Industrial Vision / Method

The tables below are design and method, not measured results. Thresholds are agreed with the client before the pilot starts and go into the acceptance document.

Edge vs cloud: where does inference run?

SituationWhereWhy
Tight per-image latency; the line cannot waitEdge inferenceInference completes locally, no cloud round-trip
Must keep running offlineEdge inferenceModel and verdict both on the device
Data must not leave the plantEdge inferenceImages and results stay on site
Batch offline analysis, cross-plant comparisonCloudAggregate analysis needs no real-time path

Camera & module selection criteria

No single vendor presumed; selection conclusions rest on pilot measurements, not vendor datasheets.

DimensionWhat the pilot checks
Resolution & field of viewWhether the smallest defect size is resolvable
Frame rate vs line speedWhether every part gets a sharp image
Lighting & calibrationWhether capture lighting is calibrated, and what happens outside the calibrated range
Interfaces & supply-chain availabilityWhether interfaces fit the existing line and lead times are checkable

Pilot plan: four gates

Each step gates the next; pilot data decides whether the scheme continues.

GateEntry conditionExit standard
1. Dataset selectionDefect categories and labelling rules fixedTrain / validation / test sets separated; test set sealed
2. Offline validationDataset selectedThree metrics meet the acceptance thresholds
3. Edge-device pilotOffline validation passedEdge-measured latency and stability meet the bar
4. Site acceptancePilot passedEach acceptance-document item checked off

Failure modes & handling

Design, not measured results.

Failure modeHandling
False positivesRejected in human review and logged; the review record feeds back into the labelling rules
MissesFound by sampling and review, added as new samples, triggering offline re-testing
Lighting changesCapture is calibrated for lighting first; batches outside the calibrated range pause automatic judgement and go to human review
New defect typesNever forced into an existing class; marked as unseen and routed to human review

Acceptance metrics

Thresholds are agreed with the client before the pilot starts and go into the acceptance document.

MetricHow it is measuredThreshold
Detection accuracyShare of defects correctly detected on the test setAgreed before the pilot
False-positive rateShare of good parts judged defectiveSame
Miss rateShare of defects not detectedSame
Per-image inference latencyEnd-to-end inference time per imageSame
Continuous-run stabilityRecord of fault-free operation over a continuous runSame

NEXT STEP

Start with the priority problem, not a generic technology demo

Prefer the weekly judgment series first? Control Brief

See whether your industrial AI scenario fits a pilot

Industrial AI starts with pilot validation: prove effect and stability at small scale first. Assess an industrial AI scenario and we will suggest a first step based on your equipment and site.