Back to work

04 / Data Science / May 2026

Defect Detection

Computer vision defect detection workflow for classifying visual anomalies and improving inspection consistency.

01

Overview

A computer vision project for identifying visual defects and turning inspection logic into a usable web-facing workflow.

02

Challenge

Manual quality checks can be inconsistent, especially when visual standards are subtle or repeated across many examples.

03

Outcome

The result is a compact data science product that connects machine learning experimentation with a frontend experience people can actually use.

Project background

Why this project exists

Defect Detection began as a machine learning exploration, but the more interesting product question is how model output becomes usable. A classification model can be technically promising, yet still fail if the review flow is unclear.

The project frames computer vision as inspection support. The interface should make the prediction visible, but also leave room for human judgment, quality review, and practical decision making.

Build notes

How it was shaped

01

Prepared visual examples and classification logic for defect-oriented model exploration.

02

Designed a simple interaction surface for uploading or reviewing inspection cases.

03

Framed model output as decision support instead of a black-box verdict.

© Contact コンタクト

Let's build something useful, impactful and beautiful.

Jakarta, --°C · --:--:--

© 2026 Rafli Ardiansyah.