Unsupervised AI
Anomaly Detection — Unsupervised Surface AI
A learn-from-good model that flags surface deviations without pre-defining or labelling every defect type — porosity, dents, contamination, and surface irregularities. Ideal for variable castings and parts where defects are unpredictable.
Good-parts-only trainingNo defect labels neededSurface-deviation mapsEdge deployment
Anomaly Detection
Key Features
No Defect Labels
Train on good parts only. The model flags anything that deviates — even defects you have never seen before.
Heatmap Localization
Returns a pixel-level deviation map so operators see exactly where and why a part was flagged.
Great for Castings
Handles natural part-to-part variation while still catching real surface and structural anomalies.
Use Cases
- Casting porosity and surface irregularity detection
- Contamination and foreign-object detection on variable surfaces
- Novel-defect detection where defect catalogue is incomplete
More AI Machine Vision
Core Vision Engine
CT Garud
AI defect detection & segmentation — 10+ defect types in one frame, <50ms per inference.
Explore →
No-Code Platform
CT Flow Builder
Drag-and-drop AI inspection pipelines. No programming. Multi-camera. Go live in hours.
Explore →
OCR / OCV
CT OCR / OCV
Retrainable OCR for batch codes, expiry dates, CIJ, laser & emboss prints at up to 400 PPM.
Explore →
/ Get Started
See Anomaly Detection on Your Line
Tell us what you need it to catch. We'll assess feasibility, and trial it on your actual parts where it stacks up.