Bring qualified cameras already in the plant or choose supported new ones.
Use a supported factory-owned edge device. Production inference always runs here.
Train in a supported cloud account your company owns. Keep your images, models and inspection IP.
Add unlimited use cases within the licensed factory.
Factory-owned AI. Factory-wide value.
Start sooner. Add inspections, not software licenses. Keep the AI you build.
A working inspection is the beginning, not the limit. Your team can train, deploy and update models through a no-code workflow. Each new use case builds on the same platform and factory software license.
Build the business case for your factory
One inspection, made visible
See an oil-leak inspection decision.
Make the condition visible.
This camera view shows a fluorescent region on the housing.
Locate what needs review.
The orange box marks the region flagged by the model.
Connect evidence to a decision.
The screen displays FAIL. The plant defines the follow-up action.
Prove it on your line. The displayed confidence is not measured inspection accuracy. Validate missed defects, false rejects and capture-to-response timing on representative production parts.
Plant-owned inspection workflow
-
01
Capture
Collect representative images where the quality challenge occurs.
-
02
Define
Plant experts identify the product, region, defect, severity, and decision.
-
03
Train
Train a customer-owned model in your company's cloud account through the platform's no-code workflow.
-
04
Deploy
Run the trained model on a supported factory-owned edge device. Production inference always happens at the edge for real-time detection.
-
05
Adapt and repeat
Add or retire defect classes, retrain, redeploy, and move to the next inspection.
Implementation support remains available when needed. Plant teams simply do not have to return to an outside specialist for every inspection change.
Inspection loop
Watch plant expertise become a live, adaptable inspection.
The camera supplies the frame. The plant expert defines what matters. United Vision carries that decision to the edge and makes feedback part of the next improvement.
Capture
Collect representative images where the quality challenge occurs.
Start with production evidence from the line and condition where the inspection needs to work.
Define
Plant experts identify the product, region, defect, severity, and decision.
The people closest to the work define what the inspection needs to recognize and how the result should be used.
Train
Train your model in your cloud, without writing AI software.
Use representative visual evidence to train and refine the inspection for the actual production condition.
Deploy
Run the trained model on a supported factory-owned edge device. Production inference always happens at the edge for real-time detection.
Bring the trained inspection back to the line so incoming images can support the configured operating response.
Adapt and repeat
Add or retire defect classes, retrain, redeploy, and move to the next inspection.
Keep the inspection current as products and conditions change, then apply the same plant-owned workflow to another use case.
Platform
See the work behind a live inspection.



Training · existing product screen
Turn production examples into a model to validate.
Label representative production images, train a candidate model and validate it against held-out evidence before release. Review new versions as production changes.
View the full training screenDetection · existing product screen
Review the flagged region and the inspection decision.
The product screen shows an oil-leak inspection and its displayed decision. Model confidence is not measured inspection accuracy.
Sub-50ms verdicts depend on the supported edge setup, model and image conditions. Qualify capture-to-action timing for the actual line; a verdict time is not the complete response time.
View the full detection screenBefore deployment · existing training screen
Review the model before approving its release.
This screen shows training options and iteration history. The displayed training metrics do not establish production accuracy or show a completed deployment. Validate the model on held-out production evidence, then qualify the target edge device and release process.
View the full training-history screenAdministration · illustrative view
Agree who can update and approve an inspection.
This illustration outlines the management workflow. Confirm user roles, version records and release responsibilities for the proposed deployment.
View the full illustrationInspection capabilities
Eight inspection capabilities. Built around the jobs already on your floor.
From visible surface flaws to hidden inclusions, identification, assembly, movement, and safety, United Vision gives plant teams a consistent way to turn camera evidence into operating decisions.
Surface & X-ray
Define visible surface defects or X-ray-visible conditions, then validate them on representative images. Surface appearance does not establish internal soundness.
- Visible-surface segmentation
- Analysis of compatible X-ray images
- Customer-defined severity thresholds
Dimensional gauging
Measure lengths, gaps, radii, hole positions, and tolerance bands without contact.
- Sub-pixel edge measurement
- Nominal-spec comparison
- Out-of-tolerance flagging
OCR & characters
Read printed, etched, embossed, and dot-peened serial numbers, date codes, and lot IDs.
- Low-contrast character reading
- Format and value verification
- Mismatch and legibility checks
Assembly & presence
Confirm that required components are present, correctly placed, and properly oriented.
- Presence and absence checks
- Position and orientation checks
- Variant-specific inspection recipes
Label & barcode
Check label appearance and read supported printed fields or codes. Expected-value matching needs a qualified reference; reading is separate from standards-based print-quality grading.
- Placement, skew, and clarity
- Printed-text cross-checks
- Record and package-association checks
Anomaly detection
Establish a visual baseline from good parts and flag deviations that do not belong.
- Good-part baseline training
- Heat-mapped deviations
- Expert feedback for retraining
Count & sort
Count and classify mixed parts, then connect the decision to the configured sort or reject path.
- Type, grade, and defect classes
- Reject and diverter triggers
- Live counts and yield visibility
Safety-zone monitoring
Monitor configured zone events as an advisory aid. Safety-rated protection and regulatory compliance require separate systems and assessment.
- Zone and machine-envelope watch
- Site-defined zones and conditions
- Time-stamped event evidence
Compare
Compare your inspection options.
Understand when rules, AI or a hybrid fit your inspection. If you are already evaluating suppliers, open a direct comparison below.
AI vs traditional machine vision
See where rules, learned models or a hybrid fit, and what to validate before choosing.
Read the methods guideSupplier comparisons
-
United Vision vs Cognex
Compare installed systems, camera reuse and the path to additional factory inspections.
Read comparison -
United Vision vs KEYENCE
Compare VS and IV4-H station configurations with a factory software platform.
Read comparison -
United Vision vs Overview
Compare Spark and OVX Core, training location, hardware choices and software scope.
Read comparison -
United Vision vs Instrumental
Compare visual inspection needs with electronics investigation and manufacturing data workflows.
Read comparison
Use cases
The inspections are different. Plant-owned Vision AI is the constant.
Start with the production decision you need to protect. Then measure the value of improving it, from fewer defect escapes and less rework to more consistent inspection.
Oil-leak detection
Find the fluorescent evidence before the housing moves downstream.
- Part enters the ultraviolet inspection cell
- The inspection view reveals fluorescent leak evidence
- Leak evidence triggers a visible fail decision
- Production decision
- Review flagged leak evidence before releasing the housing downstream.
- Value to measure
- Confirmed leaks, false rejects and downstream rework cost.
Your IP
Your images. Your inspection intelligence. Your IP.
Your company owns its trained models, plant-floor images and customer-specific inspection IP. The production knowledge captured in each model becomes an asset your team can build on, not knowledge confined to an outside developer's project.
Training runs in a supported cloud account your company chooses and owns. United Vision does not use customer IP to train or improve shared, general, or third-party AI models.
Explore model and data ownershipInsights
Practical guidance for building plant-owned visual inspection capability.

Operations
Why visual inspection use cases stay on the backlog
Understand the barriers that keep known inspection opportunities from reaching production. Read why inspection use cases stall
Economics
One-off projects versus a factory-wide platform
Evaluate the delivery model and economics of the first inspection and the tenth. Compare project and platform models
Plant ownership
What self-service Vision AI actually means
See the complete plant-controlled lifecycle from representative examples to production updates. Understand self-service Vision AIFAQ
What plant teams ask about AI visual inspection.
Direct answers about manufacturing inspection, self-service, cameras, deployment, factory licensing, and customer ownership.
What is AI visual inspection for manufacturing?
AI visual inspection uses models trained on representative production images to identify defined defects, conditions, objects, measurements, text, or anomalies. It can support manufacturing quality checks where visual conditions vary, while plant teams remain responsible for defining, validating, and acting on the inspection.
What is United Vision?
United Vision is hardware- and cloud-agnostic Vision AI for manufacturing. Plants can bring supported cameras and edge devices, keep production inference at the edge, run United Vision cloud services in a supported cloud account they choose and own, and add unlimited use cases under one factory license.
How is United Vision different from a one-off machine-vision project?
United Vision gives plant teams control of the inspection-model lifecycle and supports unlimited use cases under one factory license. Teams can capture examples, train, deploy, update, and repeat without making every new inspection another United Vision software purchase or specialist-led model project.
What does self-service mean?
Plant teams can control the working model lifecycle largely on their own: capture production examples, define and train the inspection, deploy it to a supported factory-owned edge device, retrain it as conditions change, and repeat the process for additional use cases. No data scientists are required for routine use, and implementation support remains available.
Can United Vision use our cameras and edge devices?
United Vision supports qualified customer-owned cameras and edge devices. The equipment, image quality, interfaces, line conditions and inspection requirements are evaluated together; compatibility is not universal and no mandatory United Vision camera or edge appliance is required.
What does one factory, one price mean?
United Vision uses a recurring factory software license with unlimited use cases within the licensed factory. Adding an inspection does not require another United Vision software license for that use case. Hardware, customer-cloud consumption, integration, validation and services are scoped separately. Confirm the factory boundary, renewal terms and deployment scope in the proposal.
Does self-service mean no implementation work?
No. Camera placement, lighting, integration, validation, and operational response still need to fit the use case. Self-service means the plant is not forced to depend on an outside specialist for every model change or new inspection.
What types of inspection can the platform support?
The current capability set includes surface and X-ray inspection, dimensional gauging, OCR and character reading, assembly and presence checks, label and barcode verification, anomaly detection, count and sort workflows, and configured safety monitoring.
Who owns the plant images and inspection work?
The customer owns its trained models, plant-floor images and customer-specific inspection IP. Training runs in a supported cloud account the customer chooses and owns. This does not transfer ownership of the United Vision platform software; licensing, access and any export arrangements are defined in the agreement.
How should a plant begin?
Start with a high-value inspection where the condition, visual evidence, and operating response can be clearly defined. Prove that workflow, then decide where expansion makes sense.
How should we calculate ROI for AI visual inspection?
Establish a baseline for the inspection: defect escapes, scrap, rework, false rejects and the current effort required. Estimate improvements using a validated trial, then compare the cash benefit over the same period with software, equipment, cloud, integration, validation and operating costs. Count released staff time as capacity unless it produces an actual cash saving, and avoid counting the same improvement twice. United Vision's three-year business-case worksheet lets you enter your own assumptions; it does not promise a particular return.
What belongs in the total cost of ownership?
Include the recurring factory software license, cameras and lighting, edge compute, cloud consumption, integration, implementation, validation, training and ongoing maintenance. Reuse of suitable equipment can avoid a purchase, but qualification still matters. Budget for new use cases and production changes across the period you are evaluating, not just the first installation.
What changes when we add another inspection?
Within the licensed factory, the additional use case is covered by the same United Vision software license. Your team still defines the condition, supplies representative images and validates the result. A new inspection may require additional imaging, edge capacity, cloud usage, integration or services. The advantage is removing a separate software-license decision for each use case, not claiming every deployment is free.
How do engineering teams work with operations teams using United Vision?
Operations and quality teams bring process knowledge, define acceptable results and provide production examples. Engineering teams turn those requirements into a working inspection by qualifying cameras, lighting, edge devices and line interfaces; IT governs the customer-owned cloud environment and access. The teams use United Vision's no-code workflow to train and validate models, deploy them to the factory edge and update them as conditions change. They agree acceptance criteria, operating responses and release responsibilities together, with implementation partners supporting integration, maintenance and rollout where needed.
Does United Vision require proprietary cameras, edge devices or a vendor-owned cloud?
No. United Vision is hardware- and cloud-agnostic within supported configurations. Bring qualified cameras and edge devices, keep production inference on the edge device at the factory, and run United Vision cloud services in a supported cloud account chosen and owned by your company.
Start with the inspection you need to improve.
Tell us what you need to detect, where it appears and what should happen next. Ask a technical question or request a focused 45-minute working session. A finished specification is not required.