MANUFACTURING
Surface Defect Detection
Surface Defect Detection. What it takes to run it.
High-speed identification of scratches, dents, and cracks.
// 01 · The question
“Does this part ship, get reworked, or get scrapped?”
// 02 · Why it is hard
Defects are small, low-contrast and photographed on reflective surfaces where a lighting artefact looks exactly like a flaw. And the good parts vastly outnumber the bad, so the training set is unbalanced by construction.
// 03 · What the labels have to be
Instance segmentation
Decided before the first frame
Disposition is a severity call, and severity is size and location on the part. A scratch of a certain length in a cosmetic zone is a different decision from the same scratch on a hidden face, and only a mask carries the length.
This is the decision that is expensive to reverse. The geometry has to match what the answer contains, and finding out it does not means labelling the set a second time.
// 04 · How we run it
Three parts of one loop, on this job.
LexAnnotate
Defect Annotation
Pixel-level defect annotation for surface scratches, cracks, dents, weld faults, and assembly misalignments. Multi-sensor alignment for multi-camera inspection systems.
LexInsight
Line Performance
Connect defect detections to production data - batch numbers, machine IDs, shift schedules. Identify defect patterns across lines and predict maintenance needs before unplanned downtime.
LexAlert
Quality Alerts
Monitor defect detection accuracy across changing production conditions. Detect when lighting shifts, new materials, or equipment aging degrades model performance. Auto-retrain to maintain 99%+ accuracy.
// 05 · What breaks it after launch
Quality tightens the threshold and the images do not change at all. Parts the model was taught to pass are now failures, so it stays confident and starts being wrong, and no confidence monitor can detect it.
It is not the only one that can get this use case, it is the one that usually gets it first. All ten conditions.
// 06 · In manufacturing
Defect Detection at Scale
The Challenge
Vision models performed well in pilots but lost accuracy as factory conditions shifted.
Our Execution
Built pixel-level annotated datasets and multi-sensor alignment for reliable defect detection.
The Result
Maintained 99%+ defect detection accuracy across changing production environments. Reduced unplanned downtime.
99%+
Accuracy maintained in production