Amazon Rekognition — Image and video analysis API for face detection, object recognition, content moderation, and more.
What is Amazon Rekognition?
Amazon Rekognition is a computer vision service that analyzes images and videos to extract insights without requiring machine learning expertise. It’s a fully managed API that can detect objects, faces, text, scenes, and activities — plus moderate content for safety.
Key Point: Rekognition is “vision as a service” — upload an image or video, get JSON results with labels, bounding boxes, confidence scores.
Key Features
| Feature | Description |
|---|---|
| Object & Scene Detection | Identify thousands of objects and scenes (cars, trees, buildings, etc.) |
| Face Analysis | Detect faces, compare faces for similarity, search faces in collections |
| Face Recognition | Build face databases and identify known faces |
| Text Detection (OCR) | Extract text from images (street signs, documents, products) |
| Content Moderation | Detect unsafe or inappropriate content (violence, nudity, weapons) |
| Celebrity Recognition | Identify celebrities in images and videos |
| Custom Labels | Train custom models for your specific objects (industrial parts, products) |
| Video Analysis | Analyze videos for persons, paths, text, and activities |
Use Cases
Content Moderation
Social media platforms, marketplaces, and user-generated content sites use Rekognition to automatically detect and flag inappropriate images before they go live.
Face-Based Authentication
Verify users by comparing live selfies against stored reference images (banking, access control, KYC).
Document Processing
Extract text from physical documents, IDs, receipts, or forms for digitization.
Surveillance & Security
Detect persons, track movement paths, and identify faces in video streams (retail, facilities, public safety).
Media & Entertainment
Automatically tag photos and videos with metadata for searchability and organization.
How It Works
1. Send Image/Video: Upload to S3 or provide base64-encoded data
2. API Call: Call appropriate API (DetectLabels, DetectFaces, RecognizeCelebrities, etc.)
3. Receive Results: Get JSON with labels, bounding boxes, confidence scores
{
"Labels": [
{"Name": "Person", "Confidence": 99.5},
{"Name": "Car", "Confidence": 98.2}
],
"Faces": [
{"BoundingBox": {"Width": 0.2, "Height": 0.3}, "Confidence": 99.9}
]
}4. Use Results: Integrate into your application logic
Pricing & Free Tier
| Aspect | Details |
|---|---|
| Free Tier (first 12 months) | 5,000 images/month for DetectLabels, DetectFaces, CompareFaces, RecognizeCelebrities |
| Image Analysis | $0.001 per image (first 1M images/month for common APIs) |
| Video Analysis | $0.10 per minute for label detection, face search |
| Face Storage | $0.01 per 1,000 faces stored in collections |
| Custom Labels | Training: $0.04/image · Inference: $0.001/image |
Cost Tip: Use free tier for testing and development. Most operations cost fractions of a cent.
⚠️ Pricing Disclaimer: AWS pricing is subject to change. Always verify current pricing at the official Amazon Rekognition pricing page.
When to Use Rekognition
| Use | Don’t Use |
|---|---|
| Quick image/video analysis | Custom computer vision models |
| Face detection & comparison | Biometric identification at scale |
| Content moderation | Real-time video processing under 100ms |
| OCR from photos | Complex document layout understanding (use Textract) |
Rekognition vs Textract
| Aspect | Rekognition | Textract |
|---|---|---|
| Focus | Image/video content | Document structure |
| OCR | Basic text detection | Advanced (tables, forms, key-value) |
| Best For | Photos, scenes, moderation | Documents, invoices, forms |
Important Notes
- People pathing discontinued (October 31, 2025): Path tracking for persons in video is no longer available
- Custom Labels: Available for industry-specific use cases
TL;DR
- Rekognition = Computer vision API for images and video
- Features: Objects, faces, text, moderation, celebrities, custom labels
- Free Tier: 5,000 images/month for first 12 months
- Pricing: ~$0.001 per image for most operations
- Best for: Content analysis, moderation, face detection, OCR from photos
- Not for: Complex documents (use Textract), custom CV models (use SageMaker)
Resources
Amazon Rekognition Official product page and overview.
Rekognition Documentation Complete API reference and guides.
Rekognition Pricing Detailed pricing breakdown.