Amazon Rekognition

What is Amazon Rekognition?

Amazon Rekognition is an AWS computer vision service that detects objects, faces and unsafe content in images and video. It is used through an API and billed on usage.

Key capabilities of Amazon Rekognition

• Image analysis: Group 1 and Group 2 APIs for face and detection operations.
• Video analysis: Analyze streaming and stored video.
• Face liveness: Check that a live person is present, at $0.015 per check for the first 500K.
• Face metadata storage: Store face vectors and user vectors for search.
• Custom Labels: Train models to detect your own objects.
• Custom Moderation: Adapt content moderation, priced at $0.0012 per image for the first 1M.
• Image properties: Return attributes such as dominant colors, from $0.00075 per image.

How Amazon Rekognition works

You send an image or video to the Rekognition API, or train a Custom Labels model on your own data, and receive structured results such as labels, bounding boxes and confidence scores. Billing is per image processed or per video minute, and face features need stored vectors for later matching.

Who uses Amazon Rekognition?

Rekognition is used by developers and companies on AWS for identity verification, content moderation and media search. Applications decide how to act on the confidence scores that the API returns.

Amazon Rekognition pricing

Image analysis costs $0.001 per image for the first 1 million images per month, then $0.0008 for the next 4 million. Stored video label detection is $0.10 per minute and streaming events are $0.00817 per minute. New accounts get 1,000 images per month and 60 video minutes free for 12 months.

Amazon Rekognition alternatives

Alternatives include Google Cloud Vision AI, which offers image labeling and OCR, Microsoft Azure AI Vision, which adds spatial analysis and OCR, and Clarifai, which provides a full model platform.