AI-900 Practice Question: Describe features of computer vision workloads on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Object detection

  Why it's wrong here

  Object detection is a computer vision technique that identifies and localizes specific objects (e.g., cars, people, furniture) within an image by drawing bounding boxes around them. It does not interpret or transcribe the text content that appears inside those boxes, so it cannot extract printed amounts, dates, or vendor names from an expense report. While it could potentially detect the document as an object, it would never read the printed text itself, making it irrelevant for automating expense report processing.

• ✓

  OCR (Read API)

  Why this is correct

  Correct. The Read API (part of Azure AI Vision) is specifically designed to extract printed and handwritten text from images and PDFs, returning each text string along with bounding boxes and confidence scores. It performs optical character recognition (OCR) at the character, word, and line level, making it the appropriate service when the goal is to digitize the printed information on expense reports, such as line items, totals, and dates. This is the core step in automating expense report extraction.

• ✗

  Semantic segmentation

  Why it's wrong here

  Semantic segmentation is an image analysis technique that assigns a class label to every individual pixel, creating a detailed mask that partitions the image into regions like 'background', 'person', or 'signature'. It focuses on understanding the scene structure at a pixel level, not on recognizing or transcribing characters. Therefore, it cannot extract printed text from an expense report; instead, it would only tell you which pixels belong to a receipt or a table, not what the words say.

• ✗

  Image Analysis (description generation)

  Why it's wrong here

  Image Analysis in Azure AI Vision includes a description-generation feature that produces natural-language captions summarizing the general content of an image, such as 'a person holding a paper receipt'. While these captions can provide high-level context, they do not return the actual printed text strings found on the document. To extract the specific numeric amounts, vendor names, or other printed data, you must use OCR via the Read API rather than a captioning model, because the latter does not transcribe characters.

Go deeper

Related to this question