Notes on building vision you can trust.
Short essays on uncertainty, honest baselines, and why labelling less can teach a model more.
Open-Vocabulary Detection: Zero-Shot Defect Finding
Open-vocabulary detection finds defects you never trained for—describe them in plain words and get an instant inspection baseline, zero labels needed.
Active Learning Explained: Train With Fewer Labels
How active learning uses uncertainty sampling to train an inspection model with the fewest expert labels — a guide for engineers.
Concrete Crack Detection: Grading Cracks with AI
How computer vision handles concrete crack detection — measuring crack width, grading severity and routing only uncertain cracks to an engineer.
Surface Defect Detection with Computer Vision
Detect scratches, pitting and corrosion with AI surface defect detection — start from an open-vocabulary baseline with no training data.
Remote Visual Inspection in Nuclear: Cut Dose & Downtime
Remote visual inspection with AI cuts nuclear radiation dose and downtime — routing only low-confidence defect calls to your experts for review.
Computer Vision Inspection vs Manual Inspection
How computer vision inspection compares with manual inspection on speed, cost and consistency — and where an uncertainty-first baseline fits.
What Is Image Annotation? A Field Guide for Engineers
Image annotation explained for engineering inspection teams: annotation types, bounding box vs segmentation, and how to skip manual labelling.
Wind Turbine Inspection with AI: Drone Images to Defects
AI wind turbine inspection: turn drone images into ranked blade-defect maps from an open-vocabulary baseline — no training data, uncertainty-led review.
Bridge Inspection with Computer Vision: Cracks & Corrosion
How automated bridge inspection with computer vision finds cracks and corrosion in drone imagery, with uncertainty-aware review for engineers.
AI Weld Inspection: Detecting Defects with Computer Vision
How AI weld inspection finds porosity, cracks and lack of fusion in images — and why uncertainty quantification matters in safety-critical work.
Non-Destructive Testing Meets AI: Where CV Fits Into NDT
How computer vision fits into non destructive testing — which NDT methods produce images, where AI adds value, and how to keep experts in the loop.
Active Learning for Computer Vision: Cut Labelling Effort
How active learning for computer vision cuts image-labelling effort for engineering inspection teams — fewer labels, faster models, safer review.
Automated Visual Inspection: A Guide for Engineers
How automated visual inspection uses computer vision to find defects on safety-critical assets — a practical guide for nuclear, energy and infrastructure engineers.
The Truest Portrait Is a Caricature
A caricature can be a better likeness than a photograph. The same paradox explains why a handful of well-chosen images can beat a warehouse of them.
A Good Forecast Is Often Wrong
A confidence score is meaningless unless it is calibrated. What weather forecasting learned about honest probabilities, and why modern neural networks forgot it.
Customs Doesn't Open Every Suitcase
Most luggage is never inspected, and the border works anyway. Labelling image data should work the same way: spend your attention where the doubt is.