Using AI-Generated Product Image Recognition

1. What do we mean by product image recognition?

Product image recognition uses deep learning technology enabling artificial intelligence (AI) to learn by example. AI technology identifies objects, analyzes the image for objects, classifies them, and adds relevant tags, making it easier to locate your digital assets.

Machine vision technologies combine digital cameras with AI software to recognize the characteristics of product images. Product image recognition can then perform several tasks with the information collected, including:

• Labeling the content with meta-tags
• Performing image content searches
• Guiding autonomous robots to perform certain tasks
• Building smart photo libraries based on specific categories
• Targeted advertising
• Interactive media

The process makes it easier for marketing teams to add image tags for SEO and enriched product information. You can create user-defined criteria to ensure the tags used are relevant to your products.

2. Automating attributes from product images

The AI module automatically analyzes uploaded images for classification, identifying objects, deciding what they are, and then adding relevant image tags.  These tags are added based on “confidence” levels and typically use an 80% threshold, but this feature can be user-defined.

Once the tags are generated, they can be used for attribution through AI, such as color, material, size, and function. Pictures literally say a thousand words, empowering your team to quickly onboard new products and align digital assets with accurate attributes and product descriptions.

3. How Pimberly handles attribution from images

Pimberly has always understood the importance of the PIM/DAM connection. As a result, we were never a ‘legacy PIM’ because we’ve always included a DAM feature. For us, the product image recognition functionality was inevitable from a software standpoint. It was a natural progression to use images to generate product information easily.

The PIM/DAM combination helps manage both structured and unstructured data in one place. However, with product image recognition, we took it one step further, enabling eCommerce and creative teams to turn the unstructured data of images into structured data as product information. Product image recognition will only get more adept as AI improves, and we are at the forefront of technology to leverage these advancements.

Symbiotic Relationship

Suddenly you have text-based information directly related to the images, including powerful metadata and attributes key to creating effective product descriptions. Your team can enhance the accuracy between digital assets and output from AI and combine the two to generate automated product descriptions. Your texts and images are pulled into a single pipeline to generate your product attributes.

Ongoing Improvement

Because our functionality includes user-defined information, you combine machine learning with human knowledge. Your team’s hands-on experience and intimate understanding of the products provide valuable input that makes it easier for our AI to differentiate attributes when they are too similar, to increase accuracy. Also, every time your team confirms accuracy, you increase the AI’s machine-learning knowledge. As a result, the system is constantly improving its performance, further reducing the amount of time and effort required by your team.

Human Input

Product attributes can either be tangible characteristics, such as the size, shape, or color of the product, or more abstract, like the quality and branding of your products. While AI is exceptional at providing the latter, it is not as good at providing less tangible attributes. Your team can provide human input to ensure all attributes or represented properly.

Improved Customer Experience

Through product image recognition, you can also improve the customer experience. Similar product recommendations based on similar attributes help customers find what they want while also increasing the potential to see higher average basket size. When a product is out of stock, you can also become better at serving up alternatives to reduce bounces.

Machine learning can also help determine what products go well together, whether it is fashion, interior design, or computer equipment. This creates a personalized shopping experience for customers. Through common attributes, you can also provide recommendations such as “get the look” or “save with this combination” to motivate shoppers to optimize their purchases. This reflects well on your brand, helping to build trust and creating longer-lasting customer life cycles.

PIM/DAM technology leveraging image recognition and AI helps facilitate streamlined processes in the creation and use of product information.