SceneScape 2025.2 Release

Core Features

• Dynamic Camera Configuration: Support for dynamic camera configuration in Kubernetes deployments with support for GPU offload and model chaining
• REST API in Autocalibration: New REST API endpoints for autocalibration service
• Object Clustering: New object clustering service that enables analyzing properties of object clusters like shape, size, count and motion patterns
• 3D Mapping Service: New mapping service that creates 3D meshes and calibrates cameras from camera inputs
• Geospatial Map Creation: Integrated geospatial map creation and Scene LLA configuration
• Time-Chunked Tracker: Improved tracker performance through time-chunked processing to track 100 objects at 15fps on PantherLake
• Point Cloud Map Upload: Users can now upload point cloud maps
• V2X Plugin: Readily deployable V2X plugin to convert SceneScape messages to V2X format

🔧 Improvements

• Faster Deployment: Faster kubernetes deployment along with reduced dependencies and technical debt
• Container Security: Updated services image to run as non-root user
• Documentation: Improved architecture and how-to documentation
• Simplify Manager Service: leverage docker hub hosted postgres image as database and use manager solely as web server for providing API endpoints and web UI
• OpenTelemetry Integration: Added metrics and tracing to scene controller using OpenTelemetry
• EMT OS Compatibility: Adjusted deploy.sh to run on EMT OS (ITEP-76868)

SceneScape v1.4.0

SceneScape v1.4.0 introduces new features and significant performance improvements.

Feature Enhancements:

• Volumetric regions of interest
• Scene import/export
• Enhanced tracker performance to handle 50 tracks reliably
• Switch from Percebro to DLStreamer Pipeline Server as the Video Analytics engine
• Native support for Geospatial Coordinate output

Improved Software Quality:

• Refactor build systems, remove unnecessary dependencies, and optimize image sizes
• Optimized build time - ~10 mins
• Align to containerization best practices by leveraging volumes, secrets and configs

Documentation Updates
CI/CD Improvements
Security Enhancements
Bug Fixes

SceneScape v1.3.0

Release Notes

The SceneScape v1.3.0 marks the first time the complete Intel® SceneScape codebase is available on GitHub! We are excited to partner with the community to code, contribute, and innovate with us!

New in this release

• Percebro, Scene Controller, Scene Manager and auto calibration docker images can be built independently.
• SceneScape can be deployed using third-party mosquitto broker.
• Scene Controller can simultaneously work with object detections from Percebro and Deep Learning Streamer Pipeline Server.
• Except for scene manager service at initialization, all other services can operate independent of each other.
• Better tracker scaling with increasing object categories
• Reduced docker image sizes
• Revised documentation
• Bug fixes