VisionBase Studio is one platform to build, deploy, and run automated visual inspection — classical OpenCV and deep-learning AI in the same pipeline, wired straight to your cameras, PLCs, and production line.
Acquire, process, decide, and report. No stitched-together scripts, no separate ML stack, no custom host to maintain.
Compose a sequence from 40+ components — threshold, blob, edge, barcode, OCR, measurement, AI — and tune live. No code required.
OpenCV-grade tools alongside ONNX/YOLO inference — classification, detection, segmentation, anomaly. Train models in-app, GPU-accelerated via DirectML.
Basler, The Imaging Source, Do3Think, Inovance, GigE/USB3 (Aravis), IP/RTSP, and USB webcams — plus a simulator for offline development.
OPC UA, Modbus TCP/RTU, MQTT, raw TCP, MES/SCADA and SQL — plus digital I/O and Raspberry Pi GPIO for triggers and result signalling.
SPC X-bar/R, Cp/Cpk, yield and trend analytics, full traceability, scheduled PDF reports and a live remote dashboard.
Role-based Operator / Engineer / Developer modes, login & audit, recipe versioning, and signed per-component licensing.
The same product hears a different promise depending on who's using it. Pick your role.
Write a vision component in C#/.NET with full OpenCV & ONNX access. Package a ZIP, hot-load it. The host is already built.
For developers → IntegratorConfigure cameras, logic and PLC handshakes without a custom software project. Prove once, redeploy as a recipe.
For integrators → ProductionSPC, Cp/Cpk, yield trends and traceability turn each inspection into data you can act on — fewer escapes, less scrap.
For production → OperatorA stripped-down station: live image, big pass/fail, run controls. Guided, fast, and impossible to misconfigure.
For operators →Each component feeds the next. Acquire a frame, normalise it, find features, measure, decide pass/fail — then signal the PLC, log the result, and update the dashboard. Breakpoints, live tuning, and a built-in simulator make it tunable on the bench and trustworthy on the floor.
Send us a few sample images or a short description of the inspection, and we'll show you a working sequence — classical, AI, or both.