Manufacturing & robotics / Reading guide
Compare inspection measurements, benchmark limits, and maintenance adoption without collapsing them into one success rate.
Reading prompt: Read the task, dataset, and error definitions attached to the result. An individual study is not a general accuracy figure for an industry.
Read the full topic context →On 2,042 real metal-box images captured in an unconstrained industrial environment, the best method in a peer-reviewed study achieved 10.6% false positives and 5.41% false negatives on defect localisation, against 13.02% and 8.6% for a fine-tuned VGG-16 baseline.
Machine vision is the industrial application of imaging — a combination of sensors, lenses, lighting and software that inspects, measures or identifies a part and hands a decision to an automation system — and its vendors distinguish it from the broader term computer vision, sorting cameras into line-scan, 2D area-scan and 3D categories.
Reading prompt: Compare the benchmark conditions with the claimed production use. Preserve the distinction between a preprint's argument and independently validated results.
Read the full topic context →Across nine datasets, eleven state-of-the-art models and seven metrics, a benchmark study found that models reaching 99.9% image-level AUROC on MVTecAD degrade significantly on real-world data; its own benchmark uses six datasets with real-world defects against three with laboratory-produced defects, and excludes MVTecAD.
The benchmark study argues that image-level AUROC is not accountable for the relative importance of errors, that in production a missed defective part costs significantly more than a false positive, and that test-set-based early stopping, best-epoch reporting and centre-crop augmentation inflate published results — a cost-per-error view that sits alongside, rather than against, the practitioner account that repeated false alarms are what end deployments.
Reading prompt: Ask whether the source measures use, intention, or outcomes, and whether the population and question match the other evidence.
Read the full topic context →Predictive maintenance uses equipment condition data to service a machine shortly before it fails, as distinct from schedule-based preventive and after-failure reactive maintenance; Deloitte estimates it can cut maintenance costs by up to 25% and raise uptime by 10-20%.
Surveys disagree on the direction of predictive maintenance adoption: Fluke's 2026 Censuswide survey of 600+ decision-makers in the US, UK and Germany reports a rise from 9% to 18%, while the 2025 State of Industrial Maintenance Report reports a fall from 30% to 27% — bases of 9% and 30% cannot describe the same threshold.