The Visually Similar items is a recommendation model that helps customers discover products that are visually similar to the one they are currently viewing. This model analyzes product images to find and suggest items that look similar, based on visual features. It enhances the shopping experience by showing customers other products that share a similar aesthetic, design, or style.

The Visually Similar Model uses Gemini models to process product images and identify patterns. It then compares these visual features to recommend similar items. Additionally, the model offers customization options, allowing merchants to fine-tune the recommendations according to their store’s specific needs. 

Benefits

Enhanced Product Discovery: The model helps customers easily discover products that share a similar look or design, improving the overall shopping experience and encouraging them to explore more options.

Customization Flexibility: The ability to adjust the recommendation logic provides merchants with the flexibility to fine-tune how similar products are selected. Whether you want to prioritize products from a specific collection or exclude particular variants, the model can be tailored to fit your needs.

How it works? 

The model starts by analyzing the product images. The Gemini models process these images to understand key visual characteristics, such as color, shape, patterns, and design. This helps the model identify products that visually resemble the one the customer is viewing.

Using the visual data, the Gemini models generate representations (embeddings) for each product’s image. These visual embeddings are then compared to find products with similar visual features.

Once the visually similar products are identified, they are displayed to the customer in a "Visually Similar Items" section on the product page. 

Recommended Placements

We recommend placing the Visually Similar Items Model on the Product page.