Open-Source Machine Vision for Production Lines

How does an open platform differ from standalone OpenCV or YOLO libraries?

Libraries such as OpenCV, PyTorch, and YOLO handle vision algorithms alone, while open tools like CVAT and Label Studio cover image annotation only. Engineering teams still have to build line triggering, model versioning, operator screens, and ongoing maintenance from scratch.

Robopipe connects all four stages in one open pipeline with no black box. Training configurations are plain JSON files you can inspect or override key by key, datasets can be exported at any time, and you take your trained models with you. Read the stage-by-stage walkthrough in how the Robopipe pipeline runs from capture to inference and compare developer tooling in Roboflow alternatives for manufacturing teams.

How do labelling, training, and model versioning work?

Line staff sort thumbnails into pass and fail piles or draw boxes around defect classes. Automatic rotation, exposure, and blur augmentation makes 100 labelled frames typically do the work of 1,000.

Training runs on your GPU or on the line controller in 15 to 40 minutes. Every run logs precision and recall on a held-out set, versions datasets, labels, configs, and weights, supports one-click rollback, and compares two models live on one stream. Pro and Enterprise add auto annotation in Robopipe Vision.

When should we self-host and when should we choose a managed plan?

Self-hosting suits engineering teams that want full stack control on their own compute with GitHub community support. When a plant needs sealed IP67 cameras, a floor dashboard, an AI PLC with Modbus and EtherCAT outputs, on-site mounting, and ongoing service without tying up internal engineers, our paid rentals cover hardware and upkeep.

Frequently asked questions

Where are the source code and documentation hosted?

The source code lives under the robopipe organisation on GitHub and documentation is published at robopipe.gitbook.io/doc.

Does the open-source edition cap cameras or products?

No. Cameras, sites, and inspected products are unlimited in both the open-source edition and all paid plans.

Is AI auto annotation included in the free open-source version?

No. Automatic AI image annotation in Robopipe Vision is included only on the Pro and Enterprise plans.

Do production images stay inside our plant?

Yes. Raw frames land on local storage and nothing leaves your infrastructure unless you export it.

How can we discuss a line deployment with your team?

Book a 30-minute demo on our contact page, email jan.jelinek@robopipe.io, or call +420 728 488 116. We reply within one business day.