Skip to main content
CerebraTech AI
/work/plastic-injection-defect-detection
Internal testOur own lab · parts from a real line

Detecting sink marks and flash on injection-moulded plastic parts

Dark, glossy parts — the condition that makes vision systems fail most often — so we designed the LED bar lighting and the camera angle before training any model.

1600 × 900

Starting situation

Dark, glossy injection-moulded parts have to be checked for sink marks and flash before shipping — work that was previously done entirely by eye.

Constraints on site

Dark glossy parts reflect light, the single most common reason defect-inspection vision fails, because the reflection hides shallow defects.

What we actually did

We designed the LED bar lighting and camera angle for the glossy surface before training anything, which removed the reflections hiding the defects, then trained a custom YOLO model and ran it through TensorRT on a Jetson Orin Nano Super.

Canonical Evidence Registry

EV-2026-005Reproducibility grade Cin-development

Vision inspection acceptance benchmark plan

Dataset class: internal-approved · Industrial camera and LED bar; final line hardware pending assessment · Custom detector; runtime version pending measured evaluation

Assumptions

  • A publishable result requires a versioned test set with class balance, lighting and line-speed notes.
  • Until measurement is complete, this record is a benchmark plan rather than performance proof.

Benchmark status: planned · precision, recall, false-positive rate

Want to talk about a problem like this in your plant?

A free 30-minute call. We will tell you straight if your problem is not a good fit for AI yet.

Talk to us for 30 minutes firstSee all work

This site uses very few cookies

We use only the cookies necessary to remember your language choice and save this consent preference (your chosen theme is remembered via browser local storage, not a cookie). Our analytics tool uses no cookies and collects no personal data.

Read the cookie policy