Amazon Rekognition
- What it is
- Computer vision for images and video — no ML expertise required.
- What it detects
- Objects, scenes, and activities; text in images (OCR of signs and labels); celebrities.
- Faces — detection, comparison, and search against a face collection; age range, emotion, and attributes.
- Content moderation — flags unsafe or inappropriate imagery.
- Custom Labels — train it to recognize your own objects (a specific machine part, a logo) from a small labeled set.
- Face Liveness for anti-spoofing during identity verification.
- Use cases
- Media asset tagging, user-generated content moderation, identity verification, retail analytics.
- Not to be confused with
- Amazon Textract — structured extraction from documents (forms, tables, key-value pairs). Rekognition’s text detection reads incidental text in a photo; it does not understand document structure.