Amazon Rekognition

Machine Learning

Pre-trained computer vision API for finding objects, faces, text, and unsafe content in images and video.

Verified against AWS documentation · 2026-08-24

Key facts

  1. 01Pre-trained models behind an API — no training data, no model hosting; every result comes back with a confidence score you threshold yourself
  2. 02Image APIs are synchronous (call, get answer); Video APIs are async jobs — Start* → SNS notification → Get*, over video sitting in S3
  3. 03Face collections store face vectors, not photos — CompareFaces is 1:1, IndexFaces + SearchFacesByImage is 1:many
  4. 04Limits: JPEG/PNG only, 5 MB as raw bytes vs 15 MB from S3, stored video max 10 GB or 6 hours, 20 concurrent video jobs per account
  5. 05Streaming Video Analysis and Bulk Image Analysis closed to new accounts as of 2026-04-30 — existing users keep access, new builds use the Image APIs on extracted frames

Exam triggers

Phrases and scenarios that point here.

  • "Detect objects/faces/text in images", "no ML expertise required", "pre-trained model"
  • Filter explicit or suggestive user uploads → DetectModerationLabels
  • Find every photo containing a specific person → face collection + face search, not repeated CompareFaces
  • Text in a photo (sign, jersey, screenshot) → Rekognition; text in a scanned form or invoice → Textract

Not this service

  • Documents, forms, tables, handwriting → Amazon Textract
  • Objects the built-in label set doesn't know → Rekognition Custom Labels, or SageMaker if you need control over the model
  • Speech inside the video → Amazon Transcribe

Use cases

  1. 01Screening profile photos at signup — S3 upload event triggers DetectModerationLabels, high confidence auto-rejects, borderline scores go to a human review queue instead of a guess
  2. 02Making a photo archive searchable by person — index each face once, then "show me every shot with this person" is a single search call rather than a full rescan
  3. 03Blurring faces and plates before CCTV footage lands in long-term storage, using the returned bounding boxes so the archived copy is already privacy-safe
  4. 04Reading text off things people photograph rather than scan — delivery labels, meter readings, membership cards, shelf tags
  5. 05Cataloguing a media library so editors search "stadium, crowd, night" instead of scrubbing through timelines

How it's charged

  1. 01Billed per image per API — running label detection and moderation on the same photo counts as two images
  2. 02Image APIs split into Group 1 (face compare/index/search) and Group 2 (detect labels, faces, text, moderation, celebrities); volume tiers are counted separately for each group, so usage in one doesn't discount the other
  3. 03Face metadata storage is its own monthly line item, charged per stored face vector and user vector and pro-rated
  4. 04Video is billed per minute of video analyzed, per feature — labels plus faces on one clip bills the same minutes twice
  5. 05Core APIs have no idle cost, but a Custom Labels inference endpoint bills by the hour for as long as it's deployed, whether or not you send it anything — that's the usual bill surprise
  6. 06Free tier is 12 months from account creation, not permanent, and Image Properties is excluded from it

Commonly confused with

Amazon TextractMachine LearningAmazon ComprehendMachine LearningAmazon SageMaker AIMachine Learning

References