Question 6
Domain 2: Implement Knowledge Mining and Azure AI Search SolutionsA team has a working Azure AI Search implementation. They want to add a feature where users can find products by uploading an image of a similar product (visual search). The image is analyzed and results are returned based on visual similarity. Which additional components are required?
Correct answer: B
Explanation
Visual search needs image features converted into vectors so Azure AI Search can compare “visual similarity.” Azure OpenAI multimodal embeddings create those image embeddings, and the search index must include vector fields to store and query them.
Why each option is right or wrong
A. Azure AI Vision for image tagging + text search in the index
B. Azure OpenAI embeddings applied to images + vector fields in the index (using multimodal embeddings)
Azure AI Search can only perform similarity matching over indexed vectors, so the uploaded image must first be converted into an embedding and then stored in a vector field. In this scenario, that means using Azure OpenAI multimodal embeddings for the image and adding vector fields to the search index, because vector search in Azure AI Search is the mechanism that supports nearest-neighbor retrieval for visual similarity.
C. Custom Vision classifier + OData category filter
D. Azure AI Video Indexer + semantic search