Skip to main content

Image Extraction enrichment

Image Extraction enrichment is used to extract structured product data from images.

S
Written by SKULaunch Support
Updated over 4 months ago

Image Extraction enrichment is used to extract structured product data from images.

This agent analyses images to identify text and visual information that can be mapped to the attributes you have selected. It is designed for data enrichment, not content generation.

When to use Image Extraction enrichment

Image Extraction enrichment is a good choice when:
• Product information is visible on packaging or labels
• Specifications appear in images rather than text
• Compliance marks or symbols are shown visually

It is commonly used as a supporting enrichment method.

What Image Extraction enrichment is good at

Image Extraction enrichment is commonly used to populate:
• Weight and size information
• Ingredients or materials
• Certifications and compliance markings
• Variant indicators shown on packaging

It works best when images are clear and text is readable.

What it is not good at

Image Extraction enrichment should be avoided when:
• Images are low resolution or blurry
• Text is too small or obstructed
• Information is purely decorative or symbolic without labels

In these cases, another enrichment agent may be more reliable.

What Image Extraction enrichment uses as input

This agent relies on:
• Product images
• Packaging or label images
• Images containing visible specifications

Higher quality images lead to more accurate results.

How results should be treated

Values returned by Image Extraction enrichment are:
• Suggested attribute values
• Always reviewable and editable
• Dependent on image clarity and layout

You should review enriched values carefully before approval.

Tips for better results

• Use high resolution images where possible
• Avoid images with heavy reflections or shadows
• Enrich a small number of attributes at a time
• Treat image based enrichment as supporting evidence, not a single source of truth

Did this answer your question?