Different hardware, one deployment
Sites may use CPUs, GPUs, FPGAs, or system-on-chip devices from different vendors. Analytics should be deployable across suitable hardware without rebuilding the entire solution for each location.
Real-time analytics at the edge
Turn live video and sensor streams into timely, useful information—while choosing where processing runs and how the solution evolves.
Process streaming data close to where it is created
Cameras and sensors produce continuous streams, but sending every frame or reading to a central service can add latency, consume network capacity, and complicate privacy or connectivity requirements. Edge analytics processes relevant data near the source and can forward events or metadata for broader review and management.
Reliable software across environments
Edge systems and cloud services have different constraints. A consistent, modular deployment model helps teams operate and update analytics across both.
Models that keep pace with change
Lighting, camera angles, site layouts, and activity patterns shift over time. Models need a path for review and refinement so they remain useful in their actual deployment environment.
A configurable analytics pipeline
A streaming workflow can ingest video, audio, sensor, or network data; normalize it; apply machine-learning models; derive events; and present results for review. Teams can configure processing and the rules that turn model output into operational alerts.
Ingest and prepare
Connect supported streams and normalize data for downstream processing.
Infer and analyze
Apply models to identify events, objects, and patterns relevant to the site.
Add context
Combine events from multiple streams with operational rules for more informed alerts.
Visualize and act
Present events through dashboards, queries, reports, and notifications.
An adaptable approach to analytics
Real-world deployments benefit from the ability to tailor analytics, select appropriate infrastructure, and connect results to existing workflows.
Customize
Configure the processing pipeline, select models for the task, and tune rules to the needs of a stream or site.
Choose
Choose suitable compute—including CPUs, GPUs, FPGAs, or system-on-chip hardware—and deployment locations, from on-site systems to private, public, or hybrid cloud environments.
Integrate
Connect analytics with cameras, sensors, business systems, dashboards, and applications through integration interfaces and development tools.
From streams to site operations
A representative solution combines edge analytics with site-management tools. The exact components and placement depend on camera compatibility, connectivity, operational requirements, and the agreed deployment.
Stage 1
Streams
Compatible IP cameras and sensors provide the input data.
Stage 2
Analytics
A configurable pipeline processes streams and derives events at a suitable edge or cloud location.
Stage 3
Management
Selected metadata and events can be presented for site monitoring, notifications, queries, and reports.
Stage 4
Extension
Development tools can support site-specific pipeline and visualization requirements.
Improve models with operational feedback
A practical model-improvement process uses examples from deployment to identify gaps, evaluate changes, and update models deliberately. The process should include validation and monitoring rather than automatically changing production behavior without review.
Step 1
Observe
Review events and examples from real operating conditions, including cases where model confidence is low.
Step 2
Improve
Use reviewed examples to refine an existing model or develop a model for a new requirement.
Step 3
Evaluate
Test the candidate model against representative data and compare it with the current deployment.
Step 4
Deploy
Promote a validated model through a controlled update and continue monitoring its performance.
Deploy where each part fits best
Analytics can run on a camera-side device, an on-site gateway, an edge server, or cloud infrastructure. Distributed deployments can keep time-sensitive processing close to the camera while sending selected metadata to centralized services for management, reporting, and cross-site context.
At the site
Process streams locally for responsive analytics and site-specific requirements.
Across locations
Connect multiple sites to shared tools for configuration and operations.
In the cloud
Use cloud or private infrastructure for centralized services where appropriate.
Hardware, camera compatibility, integrations, and feature availability should be confirmed for each proposed deployment. This guide intentionally omits hardware specifications and performance figures from older materials.
Explore analytics for your environment
See how Echelon Platform use cases can be tailored to buildings, cities, factories, retail, and warehouses.