Composable inspection workflows
Connect image acquisition, alignment, scoring, classification, and decisions in a visual workflow tailored to your parts and inspection needs.
EntropticsvisionFrom inspection
Manufacturing visual inspection & quality intelligence
From inspection
to improvement.
Build visual inspections, review findings with your operators, and connect defects to the process context your quality team needs.
Inspection at the edgeInference close to your production line
People in the loopReview findings and preserve decisions
Knowledge that growsLearn from confirmed investigations
Visual inspection, connected to your process
Connect every step of quality.
Bring inspection design, shop-floor decisions, and quality engineering into a connected workflow that your team can understand and improve.
Inspection on the production line
Run inference on an on-site edge appliance. Manage the fleet by site, line, and station, with visibility into health and assigned model versions.
Operator review & feedback
Route uncertain findings to people. Confirm, correct, or request another look while preserving the original machine result.
Traceability behind every decision
Link results to images, stations, parts, revisions, and review history. Follow how a finding became a quality decision.
Evidence for root-cause analysis
Bring defect history, process signals, and part-specific failure knowledge together to rank potential causes for your engineers to investigate.
Improvement with your approval
Turn approved labels into training data and confirmed investigations into proposed diagnostic improvements. Your team reviews the changes.
Inside the product
From station setup to operator review.
From the engineer designing an inspection to the operator reviewing a part, keep the evidence close to the decision.
Inspection studio
Build the inspection your part needs.
Compose a flow, compare image views, and inspect the output of each step before deploying a version to the line.
- Connect acquisition, alignment, and analysis
- Review overlays, scores, and verdicts together
- Version the workflow as your process evolves
Camera setup & calibration
Bring the station into focus.
Connect the physical station to the inspection. Organize camera groups and align each view with its role in the workflow.
- Review camera groups and calibration status
- Align camera views with CAD regions
- Connect inspection roles to the station setup
Workflow at the edge
See how the result takes shape.
Follow the inspection graph at the station and examine what each step contributes to the result.
- Follow acquisition, alignment, scoring, and decisions
- Review node timing and scoring parameters
- Inspect the images behind each verdict
Edge inspection
Give operators a clear next step.
Bring the image, machine verdict, and supporting evidence into the station view. Uncertain results stay visible for human review.
- Inspect locally using the assigned model pack
- See image evidence alongside the verdict
- Route uncertain parts into the review loop
Review & labels
Make every correction useful.
Capture operator knowledge as traceable feedback. Keep machine and human judgments connected without replacing the original result.
- Confirm or correct the inspection outcome
- Retain reviewer identity and correction history
- Evaluate on held-out data before model promotion
How it works
From the camera to a better decision.
A continuous quality workflow, with people involved where their judgment matters.
- 01
Capture
Acquire images at the station.
- 02
Inspect
Run the assigned inspection flow.
- 03
Review
Validate findings and add feedback.
- 04
Investigate
Connect defects to process evidence.
- 05
Approve
Review proposed actions and changes.
- 06
Improve
Build on approved labels and outcomes.
Quality intelligence
Understand what’s happening around the defect.
A failed inspection is the beginning of the investigation. Connect what the camera sees to process history, engineering knowledge, and the cost of poor quality.
Investigate with evidence
Rank candidate causes using recurring defects, telemetry, and failure-mode knowledge. See supporting evidence and gaps so engineers can judge what to investigate next.
Watch for process drift
Explore historical correlations and degrading process signals. Where the data supports it, compare projected trends with engineering control limits.
Bring in part-specific knowledge
Import product definitions, control plans, and PFMEA information. Bind the right part revision to an investigation for relevant tolerances and failure modes.
Understand the cost of quality
Track scrap, rework, warranty, and appraisal costs using observed counts and rates you provide. Export the ledger for review with your team.
Ask your plant
Find answers in your own inspection and investigation history. The quality copilot links answers to supporting episodes and rules, and identifies when evidence is insufficient.
Keep improvement accountable
Use accepted investigations to propose diagnostic rule changes. Two distinct people sign off before those changes become part of the knowledge base.
From one station to your fleet
Run locally. Manage with context.
Keep inference close to the line while your quality team works across inspection records, model versions, and investigations in the platform.
- Continue local inspection through temporary platform disconnections using the last assigned pack.
- Review model versions and choose when to pull and assign updates to production stations.
- Monitor device health, heartbeat, and inference counts by site, line, and station.
Plan your deployment around your cameras, storage, identity provider, and quality process. We’ll help define the right starting point for your line.
At the edge
Image acquisition, local inference, and operator review at the production station. Buffered results reconnect with the platform when the connection returns.
In your chosen cloud
Run the platform in your own cloud environment. Bring fleet management, traceability, investigations, annotations, and training together with your operational data.
Through versioned model packs
Entroptics provides licensing and the model-pack catalog. Your team pulls versioned weights, defect definitions, and thresholds, and chooses which pack runs at each station.
What does Vision cover beyond defect detection?
Vision connects inspection design and edge deployment with operator review, traceability, root-cause investigation, process correlation, cost-of-quality tracking, and approved improvements to models and diagnostic knowledge.
What happens if a station loses its connection?
An enrolled edge device can continue local inference with its last assigned pack during a temporary platform disconnection. Results are buffered and replayed when connectivity returns. Model changes and fleet coordination depend on reconnecting to the platform.
Can operators correct the machine’s decision?
Yes. Reviewers can confirm or correct findings, and request further review. Human labels and corrections are recorded separately from the original machine verdict, so you retain the full decision history.
Does an RCA result prove what caused the defect?
RCA ranks candidate causes from the available evidence. Correlation, recurrence, and product definitions help engineers decide what to investigate; they do not establish a proven cause on their own. The product distinguishes supporting evidence from gaps and incomplete data.
How does Vision learn from our team?
Approved annotations feed versioned datasets. Evaluate trained models against a separate held-out dataset before promotion, then work with your pack publisher to package an approved artifact. Separately, confirmed investigations can support proposed diagnostic rule changes. Those changes require two-party approval, and model packs are explicitly assigned to stations.
How do we bring in product definitions?
Import control-plan CSV or QIF files to connect part revisions, tolerances, and failure-mode knowledge to an investigation. Discuss any live PLM connection with our team; available file imports do not imply that a connector for your PLM system is installed.
What should we bring to a first walkthrough?
Start with one part and inspection task. Bring representative images, examples of acceptable parts and defects, and your acceptance criteria. We can then discuss camera and edge setup, operator review, and how to evaluate the inspection on held-out examples. The PCB screens above demonstrate a workflow; suitability for your part and production conditions needs its own evaluation.
How is Vision priced?
Vision’s licensing model combines a base subscription, active edge devices, and the capability tier of your model packs. Inspection count is an operational metric rather than a per-inspection billing input. Contact us to scope your stations, inspection needs, and deployment.