Industrial Machine Vision Toolkit
From sub-pixel geometric measurement to AI-powered defect detection — VisionLab gives industrial engineers a complete, deployable machine vision solution with a Plugin SDK that integrates into any existing system.
dies / wafer
sub-pixel locate
Designed for real production environments — where accuracy, reliability, and integration speed all matter
Sub-pixel geometry, AI defect detection, and model training — everything you need in a single application. No patching together multiple libraries.
RANSAC + Devernay sub-pixel edges deliver circle-centre accuracy below 0.1 px. Reliable results even under vibration, glare, and partial occlusion.
Pure-C Plugin SDK with shared-memory IPC under 2 ms. Embed VisionLab into any C/C++ host — no OpenCV or PyTorch dependency in your codebase.
What is VisionLab?
VisionLab combines proven geometric measurement algorithms with modern AI-based defect detection in a single Qt application. It ships with a Plugin SDK so any C/C++ upper-computer can offload vision tasks to VisionLab as a sidecar process — no vision dependencies in the host.
Every algorithm is built around the mathematics: sub-pixel edge localisation, RANSAC robust estimation, and Vision Transformer feature spaces — so results hold up under real factory conditions.
Algorithms, case studies, and deep-dive technical articles
Explore the full set of geometric and AI vision tools — parameters, accuracy specs, and when to use each.
Real deployment scenarios: wafer inspection, surface defect detection, and MES integration.
Deep dives into the mathematics, implementation details, and best-practice guides behind each algorithm.
Built for Industry
VisionLab algorithms are designed to operate under real factory conditions — variable lighting, vibration, partial occlusion — where off-the-shelf solutions fall short.
In Practice
Real-world deployments powered by VisionLab
Using the VisionLab Plugin SDK to integrate sub-pixel geometric measurement into a C++ upper-computer system in under a day, with zero changes to the host codebase framework.
Deploying VisionLab's DINOv2 + PatchCore anomaly detection on a metal casting line — trained on 8 normal-sample images, no defect images required.
Using VisionLab's circle fitting algorithm to measure wafer via-hole roundness at sub-pixel accuracy (<0.1px) on a 100% full-inspection production line.
Knowledge Base
Algorithm deep-dives, integration guides, and training tutorials
VisionLab's Plugin SDK lets you add sub-pixel geometric measurement or AI defect detection to any C/C++ application — without taking on OpenCV or PyTorch as host dependencies.
A deep dive into how RANSAC, sector-centroid sampling, and Devernay sub-pixel edge detection combine to achieve circle-centre accuracy below 0.1 pixels.
PatchCore flips the defect-detection problem on its head — instead of learning what defects look like, it memorises what normal looks like. Here's the math behind it.