Cosmetic Defect Detection with Computer Vision

Summary

Cosmetic defect detection with computer vision catches visible flaws like paint scratches, chips, and dents that pass function tests but still fail on appearance, and this guide builds a working car-parts inspector end to end in Roboflow. A custom RF-DETR model detects the bumper, door, and paint_scratch, then a Workflow counts each scratch, ties it to the affected part, and outputs a pass/fail result with a repair-routing recommendation and an annotated image.

A product can work correctly and still be rejected because of how it looks. A scratch on a vehicle door, a stain on a shoe, a dent in an appliance, or a chipped coating on a metal enclosure may not immediately stop the product from functioning, but it can reduce its value, create customer complaints, and require rework.

This is where cosmetic defect detection with computer vision can help. Cameras capture product surfaces and a computer vision model can identify visible imperfections, locate them in the image, and connect the defect to the affected product or component.

In this guide, I will show you how to build a cosmetic defect detection on car parts in a manufacturing settings. The system uses a custom RF-DETR object detection model to detect car parts such as bumper, door, and cosmetic defect paint_scratch. The computer vision model identifies whether a scratch is located on a bumper or a door then the logic counts the scratches, associates each scratch with the relevant car part, generates an inspection result, recommends an action, and produces an annotated inspection image.

What Is a Cosmetic Defect?

A cosmetic defect is a visible imperfection that changes the appearance of a product. It may not prevent the product from performing its intended function, but it can affect customer acceptance, resale value, finish quality, or brand perception.

For example, a car door with a light paint scratch or chip may still open and close normally. However, it affects the car’s appearance and may req…