Question

A retail team is building a lightweight vision app in Microsoft Foundry. Store associates will upload a photo of a shelf, and the app must return a written description that identifies which products are visible and estimates how many units of each are on the shelf. Which approach should the team use to implement this feature?

Accepted Answer

Deploy a multimodal model that accepts both the image and a text prompt, then instruct it in the prompt to describe the products and estimate their quantities

Answer

Deploy a text-only generative model and paste a written summary of the shelf into the prompt

Answer

Deploy an image-generation model and ask it to produce a new picture of the shelf

Answer

Deploy an embedding model to convert the photo into a numeric vector for similarity search