Bitzu

CLF-C02 ยท AI, machine learning and analytics

Vision and document AI services

Amazon Rekognition for image and video analysis and Amazon Textract for extracting text and data from documents, versus the language services.

Key points

  • Amazon Rekognition analyzes images and video: objects and scenes (labels), faces (compare, search, liveness), text in images, unsafe content, celebrities and protective equipment.
  • Amazon Textract extracts printed and handwritten text, form key-value pairs and tables from scanned documents.
  • Contrast: these services read pixels, while Comprehend analyzes the meaning of text you already have.

Source: https://docs.aws.amazon.com/rekognition/latest/dg/what-is.html

Practice questions

Single choice

Which AWS service identifies objects and scenes, such as cars or beaches, in photos?

  1. Amazon Textract
  2. Amazon Rekognition
  3. Amazon Comprehend
  4. Amazon Polly
Show answer

Answer: B. Amazon Rekognition

Amazon Rekognition detects labels for objects, scenes and concepts in images and video. Textract is for documents, Comprehend for text meaning and Polly for speech.

Select all that apply

Which requirements can Amazon Rekognition meet?

  1. Flag inappropriate content in user-uploaded photos
  2. Compare a selfie with the photo on an ID card
  3. Extract key-value pairs from scanned invoices
  4. Check that site workers in images wear helmets
  5. Decide whether a written review is positive
Show answer

Answer: A. Flag inappropriate content in user-uploaded photos, B. Compare a selfie with the photo on an ID card, D. Check that site workers in images wear helmets

Rekognition moderates unsafe images, compares faces for identity checks and detects personal protective equipment. Pulling key-value pairs out of invoices is Textract, and judging the sentiment of a review is Comprehend.

True or false

Before Amazon Rekognition can detect common objects in images, you must train your own model on a set of labeled photos.

Show answer

Answer: False

False. Rekognition is pretrained and detects common labels out of the box with no ML expertise. Training is needed only if you choose Custom Labels for objects specific to your business, such as your own logo.

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 CLF-C02 topics in this module