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.

• ✗

  Optical Character Recognition (OCR)

  Why it's wrong here

  Optical Character Recognition (OCR) is a computer vision technique that specifically detects and transcribes printed or handwritten text pixels into machine-encoded characters. It is designed for extracting words, numbers, and phrases from documents, signs, or screenshots, but it has no ability to interpret semantic content such as objects, actions, or scene context. Therefore, while OCR could read a caption that was printed in the image, it cannot create a novel, natural-language description of the photo's content, making it unsuitable for automatic caption generation.

• ✗

  Object Detection

  Why it's wrong here

  Object detection is a vision task that localizes and categorizes individual objects within an image by drawing bounding boxes and assigning class labels (e.g., person, car, dog). Although it identifies multiple discrete entities and their positions, it outputs structured data arrays rather than a fluent, grammatical sentence that describes the overall scene or relationships between objects. Simply concatenating detected labels would produce a list, not a descriptive caption like 'a woman walking a dog in a park', so it fails to meet the app's need for human-readable, descriptive text.

• ✗

  Image Classification

  Why it's wrong here

  Image classification assigns a single class label to an entire image, typically selecting the top category from a fixed set of possible labels (for example, 'beach', 'cat', or 'sunset'). This one-label output lacks the syntactic structure and detail needed to describe multiple elements, their attributes, or their spatial arrangements, and it cannot generate a full descriptive sentence. While classification might provide a keyword that partially relates to the photo, it fundamentally cannot produce the rich, multi-faceted caption that the photo-sharing app requires.

• ✓

  Describe Image (Image Captioning)

  Why this is correct

  Describe Image, more formally known as image captioning, is a deep-learning capability that combines computer vision and natural language generation to produce a complete sentence describing the salient objects, actions, and scene context in an image. It uses an encoder-decoder architecture—typically a convolutional neural network (CNN) to extract visual features and a recurrent or transformer-based language model to generate text—to create coherent, contextually relevant captions. This feature directly matches the app's goal of automatic descriptive text generation, and it is commonly used for accessibility features such as screen readers, where a visually impaired user needs to understand what is in a photo.

Go deeper

Related to this question