Amazon Rekognition Video Overview
Amazon Rekognition Video is a fully managed machine learning (ML) service offered by Amazon Web Services (AWS) that is designed to analyze and process video content in both real-time streaming and stored video formats.
Key Functionality
Real-Time Streaming Video Events
Amazon Rekognition Video supports the analysis of live video streams, enabling the detection of objects such as people, pets, and packages in real-time. This service is characterized by its low cost and low latency, making it ideal for applications that require immediate responses. Key features include:
- Object Detection: Identifies objects in the video stream and provides bounding box coordinates, a zoomed-in image of the detected object, and the corresponding timestamp.
- Face Search: Analyzes live video streams to detect and search for faces against a repository of images, allowing for timely and actionable alerts.
Stored Video Analysis
For videos stored in Amazon S3, Amazon Rekognition Video offers a comprehensive set of analysis capabilities:
- Object, Scene, and Activity Detection: Automatically identifies thousands of objects (e.g., vehicles, pets), scenes (e.g., city, beach), and activities (e.g., delivering a package, dancing). It provides confidence scores and bounding box coordinates for common objects.
- Content Moderation: Detects inappropriate content such as nudity, violence, or weapons, and provides detailed labels with confidence scores and timestamps.
- Text Detection: Automatically detects and reads text in videos, offering options to filter text by regions of interest, word bounding box size, and confidence score.
- Celebrity Recognition: Identifies well-known persons in videos, providing the name, unique ID, and URLs to related content.
- Face Detection and Analysis: Detects up to 100 faces in a video frame, returning attributes such as gender, emotions, estimated age range, and whether the person is smiling, along with timestamps.
- Face Search: Identifies known people in a video by searching against a private repository of face images, providing similarity scores and timestamps for each match.
- Person Pathing: Tracks the movement of people in videos, providing a unique index for each person and enabling the counting of people in the video.
Additional Media Analysis Features
Amazon Rekognition Video also includes advanced media analysis features:
- Detection of Black Frames, End Credits, Shot Changes, and Color Bars: Automates the detection of these elements to facilitate content preparation, ad insertion, and the addition of ‘binge-markers’ to videos. This helps in removing unwanted segments, identifying key frames for promotional videos, and ensuring ad insertion does not disrupt the viewer experience.
Benefits and Use Cases
- Scalability: Analyze large volumes of videos stored in Amazon S3 without requiring machine learning expertise.
- Accuracy: Provides highly accurate detections with frame-accurate SMPTE timecodes and handles various video frame rate formats.
- Cost-Effective: Pay only for the minutes of video analyzed, with no minimum fees, licenses, or upfront commitments.
- Use Cases: Ideal for public safety, media and entertainment, smart homes, and any application requiring video analysis, content moderation, or automated media workflows.
In summary, Amazon Rekognition Video is a powerful tool for both real-time and stored video analysis, offering a wide range of features that make it versatile for various industries and use cases. Its ability to detect objects, faces, text, and activities, along with its content moderation and media analysis capabilities, makes it an essential service for automating and enhancing video processing tasks.