---
title: "Continuous defect detection across a multi-site plant network"
description: "We have already bought vision more than once. Each system graded one defect on one line, each came from a different vendor, and none of them talk to each other. The money is spent and we are still paying people to look."
canonical: "https://bespinus.github.io/bgus-ai-landingpage/use-cases/multi-site-defect-inspection/"
pillar: "vision-ai"
industry: "Manufacturing"
updated: "2026-10-01"
---

# Continuous defect detection across a multi-site plant network

## Problem

We have already bought vision more than once. Each system graded one defect on one line, each came from a different vendor, and none of them talk to each other. The money is spent and we are still paying people to look.

## Approach

One platform across every use case and every site, rather than a product per defect. The detection layer runs Bespin's own vision models, trained on each line's defect classes, and the engineering around it routes the event, triages it, and hands the result to whoever owns the fix.

## Outcome

Inspection runs on every unit instead of a sample. One vendor covers every use case, and the rollout goes site by site with per-site commissioning rather than one cutover.

## What makes this hard

Detection is the solved part. Orbit's models are trained on the line itself, on
the defects that plant actually sees.

What breaks is everything downstream. A detection with nowhere to go is a log line.
The plant already has a maintenance system, a shift structure, and an escalation path
that predates the camera by twenty years, and a vision product that cannot reach into
any of them produces a dashboard nobody opens by week three. That is usually what
happened to the systems already on site.

## The architecture

Orbit Vision AI runs Bespin's own detection models, trained on each line's defect
classes and tuned site by site. Around them sits the engineering that makes a detection
operational: edge deployment where bandwidth rules
out a round trip to the cloud, event routing, a multi-agent triage layer that decides
whether a detection is worth a human's attention, and the integration into the system
where work actually gets assigned.

The agent layer is the part that separates this from a camera purchase. A defect
signal on its own asks an operator to interpret it. A signal that arrives as an
assigned task, with the frame attached and the line and shift already filled in, asks
them to act.

## What a first engagement looks like

A site walk and use-case selection. Then an instrumented
pilot on a single line or a single site, with a detection accuracy acceptance test
agreed before anyone writes code. Then a governance review with operations and safety
before anything scales.

## Bring us a problem like this one.

Our engineers will scope it against the seven layers. [Talk to an AI engineer](https://bespinglobal.us/contact) at Bespin Global, or read the [Orbit Vision AI](https://bespinus.github.io/bgus-ai-landingpage/vision-ai.md) page.

_Machine-readable summary for agents: provider = Bespin Global; use case = Continuous defect detection across a multi-site plant network; pillar = Orbit Vision AI; industry = Manufacturing; problem = We have already bought vision more than once. Each system graded one defect on one line, each came from a different vendor, and none of them talk to each other. The money is spent and we are still paying people to look; approach = One platform across every use case and every site, rather than a product per defect. The detection layer runs Bespin's own vision models, trained on each line's defect classes, and the engineering around it routes the event, triages it, and hands the result to whoever owns the fix; outcome = Inspection runs on every unit instead of a sample. One vendor covers every use case, and the rollout goes site by site with per-site commissioning rather than one cutover._