Step 5: Setting Up Outputs

Time: 5–15 minutes

Your AI model is trained. Now define what happens when the camera makes a decision.

The global pass/fail

Every capture produces a single binary result: pass or fail. Even if you have 50 inspection regions doing complicated analysis, it all boils down to one answer: is this part good or bad?

This global result is what gets sent to your PLC, HMI, stack light, reject gate, or any other system.

Basic mode

Basic mode is the simplest way to configure outputs. You go to your ROIs and add rules that define what constitutes a pass.

The typical setup: All ROIs need to equal "pass" for the capture to be a global pass. If any single ROI fails, the whole part fails.

1. Navigate to the IO Block in your recipe editor
2. For each ROI, set the rule (e.g., class must equal "pass")
3. Define how rules combine: all must pass, or custom logic
4. Save

That's it. The camera now outputs pass/fail on every capture.

You can monitor the real-time state of your digital I/O pins and Ethernet/IP connections on the I/O Live Monitor page:

Advanced mode (Node-RED)

For anything beyond simple pass/fail logic, click Advanced Mode to enter Node-RED, a visual programming environment that gives you enormous flexibility.

Every image capture launches a new flow. Double-click the "All Blocks Output" node to access all the metadata from your capture as a JSON object.

Things you can build:

• Time-series analysis: "Have 7 of my last 10 parts failed? Alert the supervisor"
• Custom dashboards: Pareto charts, trend visualizations, production metrics
• Data routing: Send results to FTP, MES systems, databases
• Barcode integration: Link inspection results to part serial numbers
• Conditional image saving: Only save images when the AI detects a failure
• Email/Teams notifications: Alert when defect rates spike
• Communication protocols: RS232, RS485, MQTT, HTTP/HTTPS to external systems
• Discrete I/O: Control stack lights, reject gates, conveyors via the I/O board

Importing and exporting flows

You can import and export Node-RED flows as JSON. This lets you back up your logic, share flows between cameras, or deploy flows generated by the Auto-Integration Builder.

To access import/export, click the hamburger menu (three horizontal lines) in the top-right corner of the Node-RED editor:

To import a flow: Select Import from the menu. Paste the flow JSON into the text area, or click "select a file to import" to upload a JSON file. Choose whether to import into the current flow or a new flow, then click Import.

To export a flow: Select Export from the menu. Choose which nodes to export (selected nodes or current flow), then click Download to save as a file or Copy to clipboard to paste elsewhere.

Build flows instantly with Auto-Integration Builder

Don't learn Node-RED from scratch. The OV Auto-Integration Builder at tools.overview.ai generates production-ready Node-RED flows from plain English descriptions.

How it works:

1. Open tools.overview.ai and select Auto-Integration Builder
2. Describe what you want in plain English. For example: "Send an email when 3 failures happen in a row" or "Save fail images to an FTP server with the part serial number"
3. The AI generates a complete Node-RED flow using 50+ available node types
4. Review the flow, deploy it to your camera with one click

Supports:

• Communication protocols: MQTT, Modbus TCP, OPC-UA, HTTP/HTTPS, RS232, RS485
• Data routing: FTP, databases, MES systems, cloud storage
• Logic: Time-series analysis, conditional branching, aggregation
• Notifications: Email, Microsoft Teams, Slack, webhooks
• Hardware I/O: Stack lights, reject gates, conveyors, PLCs

You can also use Modify Mode: paste an existing flow and describe what you want changed. The builder updates the flow while preserving your existing logic.

Zero Node-RED experience needed

Even if you've never used Node-RED, the Auto-Integration Builder lets you set up complex integrations in minutes. Describe what you want, review the generated flow, and deploy.

Test IO wiring before deploying to production

Incorrect wiring on the I/O board can damage the camera, connected equipment, or both. Always verify your wiring with a multimeter and run a bench test before connecting to live production hardware.

Trigger modes

Configure how captures happen:

Trigger	Description	Best for
Manual	Button press on the camera UI	Testing and setup
Hardware (digital input)	Electrical signal from a sensor	Automated lines with proximity sensors
PLC	Command from your industrial controller	Fully automated with precise timing
Aligner	Auto-triggers when part alignment is detected	When parts arrive at unpredictable times
Interval	Captures at set time intervals	Continuous monitoring

Digital output current limits

The OV10i digital outputs have a maximum current rating. Check the hardware specifications before connecting high-power devices such as solenoids, motors, or large relays. Use an intermediate relay or driver circuit if the load exceeds the rated output current.

Deploy

1. Activate the recipe
2. Set your trigger mode
3. Run test parts and verify the pass/fail output matches expectations
4. Check edge cases, especially the hardest parts to classify
5. Monitor for the first hour to ensure consistency

Download PLC integration files

If you're integrating with a PLC, download the configuration files and sample code:

EtherNet/IP (Allen-Bradley)

File	Description
OV10i EDS File	Electronic Data Sheet for Studio 5000 (ControlLogix/CompactLogix)
Recipe Switch Routine	Ladder logic for changing recipes via PLC
Camera Trigger Routine	Ladder logic for triggering inspections and handling results

PROFINET (Siemens)

File	Description
OV10i GSDML File	Device description for TIA Portal

note

The OV10i uses the same PLC integration interface as the OV20i. These configuration files are compatible with both cameras.

tip

Import the EDS or GSDML file into your PLC programming environment before configuring the connection. The L5X routines are ready-to-use ladder logic that you can import directly into Studio 5000.

Output checklist

Before going live, confirm:

• IO rules configured (pass/fail logic matches your requirements)
• Trigger mode set (manual, hardware, PLC, aligner, or interval)
• Recipe activated
• Test parts run through (pass/fail output matches expectations)
• Edge cases tested (hardest parts classified correctly)

Your AI inspection is now live. For ongoing optimization, see Improving Your Model.