Computer vision for OEE, from detecting line events to validating stops and quality

For years, improving OEE has relied on a mix of PLC signals, sensors and manual inputs. These tools are useful, but they leave blind spots. They cannot always tell you why the line stopped, whether the stop was real or just a micro disruption, and what actually happened to product quality in that moment.

Computer vision changes that. By using cameras on the line and smart video analytics, manufacturers can see what truly happens on the shop floor and connect that insight directly to OEE, maintenance and production actions inside Fabrico.

In this article, we look at what computer vision can see on the line, how it validates stops and quality, and where it outperforms traditional sensor-based approaches for OEE improvement.

What computer vision actually sees on the line

Traditional OEE setups rely on discrete signals: machine run or stop, cycle counters, reject counters, and some manual classification of downtime. Computer vision adds a rich layer of context on top of these signals, using one or more cameras pointed at critical parts of the process.

With Fabrico, video data is combined with machine data from PLCs and other sources. The platform associates the visual stream with timeline events, so every run, stop and slowdown can be understood with picture-level detail. You can read more about how Fabrico captures and structures OEE data for leaders in operations in this article: OEE software for COOs and Operations Directors.

Events that computer vision can detect

Depending on the configuration and the process, computer vision can be trained to recognize:

• Material presence and flow, such as bottles, cartons, trays or parts moving on conveyors
• Operator interventions, such as manual stops, adjustments, changeovers or cleaning
• Blockages and jams, for example accumulation at a transfer point or star wheel
• Packaging issues, such as missing components, open flaps, skewed labels or misaligned seals
• Micro stops and hesitations that do not always register clearly on sensors
• Safety-related events, such as guards opened or emergency stops pressed

The outcome is not just more data, it is better interpretation of events that already appear on your OEE timeline. Computer vision becomes a second source of truth that confirms or corrects what your sensors report.

From line events to validated stops in OEE

Plant teams often struggle with the accuracy of downtime data. Sensors tell you that a line was stopped for 3 minutes, but not why. Operators may choose a generic code or forget to classify the stop at all. This makes it hard to trust loss breakdowns and even harder to sustain improvements.

Computer vision helps validate that a stop was real, not just a glitch, and gives visual context about what happened in that window. Inside Fabrico, this visual evidence is linked directly to the downtime record, so reliability, production and CI teams can quickly confirm the real root cause.

How vision strengthens stop detection

With computer vision connected to Fabrico, you can:

• Confirm true stops versus flow interruptions, for example a conveyor still moving but no product arriving
• Differentiate micro stops from slow cycles, so short hesitations do not disappear in average cycle time
• Verify changeover, setup and cleaning durations based on actual activity on the line
• Spot unreported stops that were masked by manual overrides or local workarounds

As a result, run and stop states become more reliable. The OEE timeline reflects real production behavior, not just what sensors think is happening. This is especially useful in complex lines where multiple machines share responsibility for a stop, or where PLC signals are incomplete or inconsistent.

Turning validated stops into actions

Accurate stop detection is only useful if it leads to action. Fabrico connects every loss to structured workflows for maintenance and operations. When computer vision highlights a recurring pattern around certain stops, teams can:

• Create or adjust standard work to reduce changeover time
• Add targeted operator checks to prevent frequent minor jams
• Trigger maintenance work requests when visual evidence suggests mechanical wear, misalignment or contamination
• Refine classification rules so future events are automatically tagged with the right loss codes

Since Fabrico includes built-in maintenance management, there is no need to jump between tools. A line stop detected and validated with video can immediately become a maintenance job, inspection task or improvement action, all tied to the right asset and time window.

Validating quality directly on the line

Sensors can count rejects, but they rarely explain why defects occur. Simple photoelectric sensors or weigh cells might tell you that a unit is bad, not what type of defect it is and how it relates to process conditions.

Computer vision adds that missing layer. It can see what is wrong with the product, not just that it is wrong.

What quality events vision can detect

With appropriate configuration, vision models on Fabrico can be trained to detect visual quality issues such as:

• Incomplete or missing components inside a pack
• Label issues, such as skew, bubbles, misplacements or missing labels
• Print quality problems, such as smudging, faded codes or date/lot errors
• Seal integrity issues that are visible at the packaging level
• Surface defects, such as scratches, dents or contamination

These detections are not limited to simple pass or fail results. Different defect types can be classified and counted, which allows teams to see quality losses by category on OEE and quality dashboards.

Connecting quality detection to OEE and actions

Within Fabrico, visual quality detections are tied to machine states and production orders. This lets you:

• See which products, SKUs, shifts or lines are associated with particular defect types
• Correlate defects with process conditions or specific machines on the line
• Quantify the real OEE impact of quality events, including rework and scrap
• Trigger targeted maintenance checks when defects hint at mechanical or alignment issues

The platform can also use video snapshots or short clips for each detected defect as evidence for root cause investigations. Teams can quickly compare normal and faulty operation, then implement corrective actions directly in maintenance and production workflows.

If you want to explore how video evidence supports ROI calculations and problem solving, see this deep dive on the topic: OEE ROI with computer vision and video replay.

Where computer vision outperforms traditional sensors

Sensors and PLC data are still essential, but they have limits when used alone. Computer vision complements them in several important ways for OEE improvement.

1. Rich context instead of binary signals

Sensors typically generate binary or numeric signals: yes or no, on or off, count up or down. They struggle with ambiguity, such as partial blockages, product misalignment or intermittent slips.

Computer vision delivers contextual information. Teams can see the shape, orientation and behavior of products and machines. This context helps them interpret sensor events correctly and avoid misclassification of stops or quality losses.

2. Faster root cause analysis

Without video, root cause analysis relies on operator memory, logbooks and assumptions. People may disagree about what happened or miss small but critical details, such as a subtle misalignment or inconsistent manual step.

With computer vision connected to Fabrico, teams can replay key moments around a stop or defect. This shortens investigations, reduces debate and focuses attention on observable evidence.

Fabrico associates each event with video segments and timestamps, so CI and maintenance teams can jump directly to the relevant part of the line history instead of reviewing hours of footage. This makes problem solving faster and more repeatable.

3. Better coverage of manual and hybrid operations

In many plants, some steps remain manual or semi automatic. Manual packing, kitting, inspection or rework areas can be major sources of variation in performance and quality, but they are hard to measure with sensors alone.

Computer vision can monitor these areas without changing the physical process. It can observe operator sequences, material handling patterns and manual inspection results, then feed that information into Fabrico as events linked to OEE and quality metrics.

This gives plant leaders a more complete view of performance across both automated and manual operations, which is crucial for continuous improvement and labor optimization.

4. Flexible deployment as products and lines change

Reconfiguring physical sensors for new formats, new SKUs or frequent changeovers can be costly and time consuming. Cameras and vision models can often be adapted using software, new detection zones or updated training data.

Fabrico is built to handle this dynamic environment. As products, lines and routings evolve, vision-based rules and event definitions can be updated without changing the underlying hardware. This is especially valuable in high mix or contract manufacturing environments where agility is essential.

How Fabrico unifies machine, vision and maintenance data

Computer vision truly delivers value when it is integrated into a broader MES and OEE environment. Fabrico brings together machine signals, vision events and built-in maintenance management in one cloud platform.

For plant and operations leaders, this means:

• A single timeline of events that combines PLC data, sensor signals and video-based detections
• Real time OEE visibility with validated stops, slowdowns and quality losses
• Direct linkage from each loss to production actions, maintenance work and improvement tasks
• Centralized access to event history, video evidence and asset records

If you are considering how OEE software should fit into your broader operations strategy, this article may help: unlock smarter manufacturing with Fabrico OEE.

When to consider computer vision for your lines

Computer vision is not required for every line or process, but it becomes very attractive when:

• You see frequent unexplained stops or micro stops that operators struggle to classify
• Quality losses are rising, but root causes are hard to observe with the naked eye
• Manual or semi automatic steps create variability that is not captured in current data
• Your team spends too much time debating what happened during incidents rather than fixing them
• You want stronger evidence to support investment cases in maintenance, automation or line redesign

In these scenarios, combining computer vision with a cloud-based MES and OEE platform like Fabrico provides a pragmatic way to increase visibility and accelerate improvement, without replacing your existing control systems.

Next steps

Computer vision for OEE is not just about adding cameras. It is about connecting what the cameras see with what your machines, operators and maintenance teams do every day. Fabrico was designed to bring these pieces together so manufacturers can understand real losses and act on them quickly.

If you want to explore how this could work on your own lines, or see examples tailored to your industry and equipment, the best next step is a conversation with the team.

• Request a demo to see Fabrico with computer vision in action and discuss your use cases
• Contact us to talk about specific lines, plants or integration scenarios