ML service that analyzes images and videos to detect and identify things automatically — no ML knowledge needed, just call the API.

One line: Give Rekognition an image or video, it tells you what is in it.

What It Can Do

Feature Example
Object & Scene Detection "This image has a dog, grass, park"
Facial Analysis Age range, gender, emotions, glasses
Face Search & Verification Match face against a database — user login via face
Celebrity Recognition Identify famous people in images
Text Detection Read text inside images
Content Moderation Flag inappropriate images or videos
Pathing Track movement of people — used in sports analysis

Content Moderation — Important Feature

Detect inappropriate, offensive, or unwanted content in images and videos.

Image/Video
     |
 Rekognition
     |
 Confidence Score  (e.g. 92% likely inappropriate)
     |
 Above threshold?  → Auto flag/block
 Below threshold?  → Send for manual review in Amazon A2I

Confidence Threshold — you set the sensitivity. Higher threshold = only very obvious content flagged. Lower = flag anything suspicious.

Amazon A2I (Augmented AI) — human review workflow for borderline cases Rekognition is not confident about.

Used in: Social media, e-commerce, broadcast media, advertising platforms.

Real Example

Instagram-like app — user uploads a photo. Before it goes live, Rekognition scans it. If it detects nudity with 90%+ confidence, it is auto-blocked. If confidence is 60-90%, it goes to a human moderator via A2I. Below 60% — published normally.

Exam One-Liners