This facility (anonymous), producing white goods components, performed quality control by sampling at line end. In-line defects accumulated into scrap cost.
Previous process
Operators tried to catch defects visually; fatigue and speed meant small surface flaws were often detected late.
Scrap rate was around 3.2% per shift; root cause analysis was always delayed.
Hype Vision quality module
Line camera video was analyzed by the model. Surface and assembly deviations were classified within 0.1–0.3 seconds.
On defect detection, the quality engineer received an instant alarm; optional PLC integration defined a line-stop signal (this site used alarms only in the first phase).
After an 8-week pilot, shift-based scrap rate fell to 2.5% (roughly 22% relative improvement).
Lessons learned
Early detection stops scrap buildup; the real savings are in response speed, not only scrap cost.
The model was fine-tuned within days on the facility’s own part samples — not a generic out-of-the-box model.
* Customer name and facility details are not shared for confidentiality. Results may vary by facility conditions.