Bitzu

AI-901 ยท Vision and image generation solutions

Choosing a vision approach and its limits

Choosing between a multimodal chat model and a dedicated vision capability, image quality limits, sensitive uses such as faces, and responsible use of generated images.

Key points

  • A multimodal chat model suits open-ended questions about an image in natural language.
  • A dedicated Azure Vision feature (captions, tags, objects, Read) suits repeatable, structured output at scale.
  • Blurry or tiny images cut accuracy either way.
  • Photos of people raise privacy and consent duties, and generated images need filters, human review and clear labeling.

Source: https://learn.microsoft.com/en-us/azure/ai-services/computer-vision/overview-image-analysis

Practice questions

Single choice

A user uploads a chart and asks, in plain words, why sales dipped in March. Which option fits best?

  1. A multimodal chat model with the image and the question
  2. Azure Vision smart crops
  3. Azure Vision people detection
  4. An image generation model
Show answer

Answer: A. A multimodal chat model with the image and the question

Open-ended reasoning over an image is what multimodal chat models do. Smart crops and people detection return fixed structures, and generation creates new pictures.

Select all that apply

Which practices support responsible use of AI-generated images?

  1. Keep the content safety filters enabled
  2. Label published images as AI-generated
  3. Have a person review images before publishing
  4. Disable filters when they slow down creative work
  5. Assume generated people never raise likeness concerns
Show answer

Answer: A. Keep the content safety filters enabled, B. Label published images as AI-generated, C. Have a person review images before publishing

Filters, labeling and human review are sound safeguards. Turning filters off or ignoring likeness concerns increases harm.

True or false

True or false: setting detail to low has no effect on how well a model reads small text in a photo.

Show answer

Answer: False

False. Low detail uses a reduced-resolution view, which can hurt recognition of small objects and text.

Practice 8 questions on this topic in Bitzu

Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.

Download on the App StoreGet it on Google Play

More AI-901 topics in this module