Create a defect detection objective and follow the instructions.

Before the system can suggest labels automatically, a minimum of 10 manually labeled defects is required. This initial labeling phase helps the model learn what constitutes a defect in your specific context.

There are two ways to perform this manual labeling:

1. Upload samples manually and assign the correct defect labels during upload.
2. Label directly from the product page by navigating to a product that contains a defect, clicking on it, and assigning the appropriate label.

Once at least 10 defects have been labeled, the system will begin providing automated label suggestions to support and accelerate the inspection process.

When marking defects, make sure to frame the errors precisely and consistently.

💡 Tip: We have an extensive guide on labelling, make sure to read it if you are new to the topic. High quality labels make or break the final model.

The more defect images you mark, the more robust the AI model will be.

Ideally, you should mark over 100 labels per defect.