Amazon Rekognition
Image recognition software
Deep learning software
- Features
- Ease of use
- Ease of management
- Quality of support
- Affordability
- Market presence
Take the quiz to check if Amazon Rekognition and its alternatives fit your requirements.
Pay-as-you-go
Free Trial
Free version unavailable
Small
Medium
Large
- Real estate and property management
- Retail and wholesale
- Information technology and software
What is Amazon Rekognition
Amazon Rekognition is a cloud-based computer vision service that provides APIs for analyzing images and videos, including object and scene detection, facial analysis, text detection, and content moderation. It is used by developers and data teams to add visual understanding features to applications such as media analysis, identity verification workflows, and safety/compliance review. The service is delivered as managed AWS APIs rather than as a model-training platform, with optional support for training custom labels for domain-specific detection.
Broad prebuilt vision APIs
Rekognition offers a wide set of managed capabilities such as label detection, face detection/analysis, face search, text detection (OCR), celebrity recognition, and moderation labels. This breadth reduces the need to assemble multiple point solutions for common vision tasks. It supports both image and video analysis through AWS services and SDKs.
Managed AWS deployment model
The product runs as a fully managed service, so teams do not need to provision GPUs, manage model serving infrastructure, or handle patching. It integrates with common AWS building blocks for storage, eventing, and access control, which can simplify production deployment for AWS-centric organizations. Usage-based pricing can align costs with variable workloads.
Custom Labels for domain needs
Rekognition Custom Labels enables training and hosting of custom object detection/classification models using customer-provided labeled images. This helps when prebuilt labels do not cover a specific domain or taxonomy. The workflow is oriented toward getting a model into production without building a full training pipeline from scratch.
AWS lock-in and portability
Rekognition is accessed through AWS APIs and IAM-based security, which can increase switching costs for organizations using other clouds or on-prem environments. Model artifacts and inference endpoints are not designed for portable deployment outside AWS. Multi-cloud architectures may require additional abstraction layers or duplicate implementations.
Limited training workflow control
Compared with platforms focused on dataset management and iterative model development, Rekognition provides less control over model architectures, training hyperparameters, and experiment tracking. Custom Labels supports a managed training experience, but it is not a full MLOps environment for complex pipelines. Teams with advanced ML requirements may need separate tooling for data versioning, labeling operations, and evaluation.
Governance and sensitive use constraints
Face-related capabilities can introduce heightened privacy, compliance, and policy requirements, depending on jurisdiction and use case. Organizations often need additional governance controls (consent management, retention policies, auditability) beyond the API itself. Some use cases may be restricted by internal policy or regulatory expectations even if technically feasible.
Plan & Pricing
Pricing model: Pay-as-you-go (usage-based)
Free tier/trial: AWS Free Tier (time-limited): Image: 1,000 images/month (Group 1) + 1,000 images/month (Group 2) free for 12 months; Image face storage: 1,000 face vectors and 1,000 user vectors/month free for 12 months; Video: 60 free minutes/month for 12 months; Custom Labels: 2 free training hours/month and 1 free inference hour/month for 12 months. (All above as stated on AWS Rekognition pricing page.)
Amazon Rekognition Image (per image, tiered by monthly volume)
- Group 1 APIs (AssociateFaces, CompareFaces, DisassociateFaces, IndexFaces, SearchFacesByImage, SearchFaces, SearchUsersByImage, SearchUsers):
- First 1,000,000 images: $0.0010 per image
- Next 4,000,000 images: $0.0008 per image
- Next 30,000,000 images: $0.0006 per image
- Above that (example tiers shown): $0.0004 per image
- Group 2 APIs (DetectFaces, DetectModerationLabels, DetectLabels, DetectText, RecognizeCelebrities, DetectProtectiveEquipment):
- First 1,000,000 images: $0.0010 per image
- Next 4,000,000 images: $0.0008 per image
- Next 30,000,000 images: $0.0006 per image
- Above that (example tiers shown): $0.00025 per image
- Image Properties (priced separately):
- First 1,000,000 images: $0.00075 per image
- Next images (example): $0.0006 per image
- Face metadata storage: $0.00001 per face metadata (face vector or user vector) per month
Amazon Rekognition Video (per minute)
- Streaming video events (per-minute example): Label Detection: $0.00817 per minute (streaming example shown)
- Stored video analysis (per-minute examples):
- Label Detection: $0.10 per minute
- Shot Detection: $0.05 per minute
- Content Moderation: $0.10 per minute
- Face metadata storage: $0.00001 per face metadata per month (applies to video collections as well)
Amazon Rekognition Custom Labels (hour-based)
- Training: $1.00 per training hour (example shown)
- Inference (model availability per hour): $4.00 per inference hour (example shown)
- Free Tier for Custom Labels: 2 training hours/month + 1 inference hour/month for first 12 months
Amazon Rekognition Face Liveness (per check, tiered)
- Example tiers (US East example):
- First 500,000 checks: $0.015 per check
- Next 2,500,000 checks: $0.0125 per check
- Next (remainder in example): $0.010 per check
Discount options / volume tiers:
- Image and Face Liveness pricing show explicit tiered volume discounts (lower per-unit rates at higher monthly volumes). See the table/examples for the tier boundaries and reduced rates.
Example costs (from AWS pricing examples on the official page):
- Processing 2.5M images with Group 2 (DetectLabels): 1M x $0.001 = $1,000; 1.5M x $0.0008 = $1,200; Total = $2,200.
- Image Properties for 2.5M images: 1M x $0.00075 = $750; 1.5M x $0.0006 = $900; Total = $1,650.
- Streaming video (Label Detection) cost example: 35,000 minutes x $0.00817/min = $285.95 per month (ongoing).
(Note: The official AWS Rekognition pricing page contains full pricing tables, region notes, and several worked examples. All figures above are taken directly from AWS Rekognition's official pricing page.)
Seller details
Amazon Web Services, Inc.
Seattle, Washington, USA
2006
Subsidiary
https://aws.amazon.com/
https://x.com/awscloud
https://www.linkedin.com/company/amazon-web-services/
