Oracle Cloud Infrastructure (OCI) is excited to announce the general availability of Streaming Video Analysis, a fully managed, GPU-accelerated analytics service delivering real-time video insights from live RTSP streams. Enterprises, security professionals, and developers can now unlock actionable information from live camera feeds—deploying advanced AI-driven analytics at scale.
What Is Streaming Video Analysis?
Streaming Video Analysis on OCI processes live videos through sophisticated AI models, detecting and tracking objects, labels, text and faces in real time. Designed for low-latency, high-performance environments, it supports high throughput and can be scaled for demanding use cases such as security monitoring, retail analytics, manufacturing, stadiums, and more.
Key Features
- Object Detection – Detects objects (e.g., people, vehicles, text, labels) and returns bounding boxes.
- Face Detection – Identifies faces with bounding box outputs.
- Weapon Detection – Identifies guns(other weapon types will be included later) across streams with bounding box outputs.
- Object Tracking – Object Tracking feature enables tracking of detected objects (currently limited to faces, guns) across video frames. By assigning a unique identifier (tracking ID) to each face, the system maintains continuity of identity as a person moves through the camera view. In case of Face Tracking, both single camera tracking and multi-camera tracking are supported.
- Single Camera Tracking: Tracks faces within the bounds of a single camera stream. This allows accurate identification and tracking of individuals as they move through the field of view of a specific camera.
- Multi-Camera Tracking: Multi-camera tracking enables identity consistency across multiple camera streams. Cameras can be logically grouped so that a face detected in one stream can be recognized in another stream. (Gun detection is not yet supported across multiple cameras.)
Technical Highlights
Performance & Recommendations
To achieve optimal results:
• Use cameras with consistent frame rates of 30FPS.
• Ensure camera resolution is above 720p.
• Maintain well-lit environments and recommended subject proximity (15–20 meters).
Flexible Deployment Models
OCI Vision Streaming Video Analysis supports both public and private network connectivity:
• Public Endpoint: Exposes streams to the internet via a public IP, suitable for flexible or temporary setups.
• Private Endpoint: Keeps streams private through OCI’s VCN for production-grade security. Integrate seamlessly with your enterprise networking using site-to-site VPN or private endpoints.
Getting Started
• Camera Setup: Connect your camera to your network (LAN or Wi-Fi), configure static IPs and required port forwarding for public endpoints, or use VPN/private routing for secure deployments.
• VCN/Subnet Configuration: Provision necessary OCI networking resources and subnets to support camera feeds and video analysis jobs.
• IAM Policies & Security: Assign precise access control with OCI policies for Vision service, networking, and resource management—via dynamic groups or all-in-one policies for streamlined administration.
• SDK Support: Manage and monitor streaming jobs through the OCI SDK, leveraging example code and integrated object storage for results.
Use Cases
• Smart Surveillance: Monitor sites in real time for personnel safety, access violations, or unusual events.
• Retail Analytics: Track customer footfall, behavior, and optimize layouts using live video.
• Manufacturing/Facilities: Detect protective gear, monitor processes, and ensure compliance dynamically.
• Sports/Entertainment: Real-time ad placement, event highlights, and identity tracking across multiple views.
For more information, see the following resources: