DeepVision Accuracy Methodology
This page explains how DeepVision accuracy is measured and lists the results from named deployments. Rates depend on the part, the defect type and the imaging setup, so each deployment is measured on that customer’s own parts and its figures are reported to that customer.
How We Measure Accuracy
Accuracy figures are derived from production deployment data collected across customer sites, not from controlled lab benchmarks. Each deployment undergoes a structured validation process:
- Baseline study — We collect good and defective parts, or images of them, from your line, checked by your quality team, before DeepVision goes live.
- Parallel run — DeepVision runs alongside existing manual inspection. Both decisions are recorded independently.
- Ground truth reconciliation — A sample of DeepVision decisions (including marginal cases) is reviewed by trained quality engineers to establish ground truth labels.
- Metrics calculation — Detection rate (true positive rate), false positive rate, and overall accuracy are calculated from the reconciled dataset.
Definition of Terms
- Detection rate (sensitivity) — percentage of actual defective units correctly identified as defective by DeepVision. Measured for each deployment; named results are listed below.
- False positive rate — percentage of non-defective units incorrectly flagged as defective. Measured for each deployment during the parallel run.
- Accuracy range explanation — the range reflects differences in defect type, material and imaging setup. Each deployment’s own rates are reported to that customer.
Results From Live Lines
- Automotive stamping — sheet-metal inspection at JBM Auto and a second automotive OEM, validated on 1,073 samples: 100% accuracy on the JBM body-in-white line and 98.2% combined accuracy across both OEM lines.
- Flat-panel boards — a 10-camera board inspection line: 98% or better (case study).
- Steel — SWIR foreign-material detection at Tata Steel: 98% or better.
Continuous Model Improvement
DeepVision models can be retrained with new production images as your products and defects change.
Request a Validation Report
Prospective customers can request a detailed validation report for their specific industry and defect types. Contact us →