Skip to content

Visual Search and Shoppable Images

Entry #632 / 700 · Chapter 13 · 13.3 Conversational, multimodal, zero-shot recall · Read in Chinese →
In one sentence

Optimising for the entry points where someone points a camera at an object or long-presses an image to find where to buy it.

Analogy Mental model

Shop windows on a street where nobody needs to come in and ask the name. A photograph of the sofa returns its name, its price, and a purchase button.

In practice What you actually do

The product must be in the merchant feed, because the shopping card that appears after a visual lookup draws its image and price from there. Supply high-resolution images with a clear subject and no text overlay, alongside real photographs of the item in use.

Where people get burned Pitfalls

A promotional banner printed across the centre of the main image violates the feed requirements and simultaneously destroys the outline a recognition model needs. It fails as marketing and as data.

This entry was written in English from the Chinese original, not machine-translated. Read the Chinese version →