2 min read

Defect Detection Vision Pipeline

Table of Contents

Overview

This project focused on automating first-pass visual inspection for repeatable part geometries using a calibrated camera setup.

Problem

Manual inspection was inconsistent during high throughput, and small defects were sometimes missed until downstream checks.

What I built

  • Controlled capture setup and calibration workflow
  • Image preprocessing for lighting normalization and edge enhancement
  • Geometric checks and tolerance rules for pass/fail decisions
  • Exported inspection snapshots for operator validation and audit

Outcome

The pipeline improved consistency of defect flagging and reduced the amount of repetitive manual checks required per batch.

Next improvements

  • Add active learning loops from operator-confirmed false positives
  • Expand support for additional part families
  • Add dashboard metrics for defect trend monitoring