Implement computer vision solutions 3.2 Design and implement multimodal understanding workflows
A retailer uses Azure Content Understanding for two needs.
Study guide skill: Implement visual understanding by configuring Azure Content Understanding in Foundry Tools to extract visual characteristics
Question B3-021
A retailer uses Azure Content Understanding for two needs. Need A: from shelf photos, detect and count products and record details such as misplaced items in defined fields. Need B: from photos of printed receipts, extract and analyze the printed text. Which approach follows the product guidance?
Select one answer.
Answer: B
Content Understanding image analyzers let you define a schema of fields, descriptions, and output types, and shelf analysis (detecting, counting, and extracting details about products) is a listed image scenario. Image analyzers aren't optimized for analysis based primarily on extracted text, so for receipts you should use a document field extraction schema instead. That rules out using one approach for both needs or swapping them.
Deep dive
How it works
Content Understanding handles images and documents with different analyzers, and its image page draws the line explicitly.
Image analyzers let you define a schema of fields, descriptions and output types, and the service returns structured data from your images. Shelf analysis and inventory management is a listed scenario: detect, count and extract specific details about retail products. Need A fits that exactly. You'd define fields such as product counts and misplaced items, and the analyzer fills them from each shelf photo.
The same page also says where image analyzers don't fit. They aren't optimized for analysis based primarily on extracted text. If your main goal is to extract and analyze text from images, use a document field extraction schema instead. A photo of a printed receipt is a text problem, so need B belongs with document analysis, which extracts essential information, fields and relationships from documents and forms.
So the design is two analyzers: image analyzer + field schema for shelves, document field extraction for receipts.
The other options, and when you'd use them
- A document field extraction schema for both: Document analysis is built for documents and forms. Detecting and counting products on shelves is the image analyzers' listed shelf-analysis scenario.
- Document schema for shelves, image analyzer for receipts: Swapped. Image analyzers aren't optimized for text-centric analysis, and shelf detection is an image scenario.
- An image analyzer for both: Fine for shelves, but receipts are text-centric. For that, the guidance is to consider a document field extraction schema.
Exam tip
Content Understanding: things in the picture (count, detect, defects) = image analyzer. Text in the picture (receipts, forms) = document field extraction.
References
Last reviewed: