Enhanced nodes include GPU optimizations for Generic Detector, advanced object filters, MQTT version specification, and external certificate support for Google Chat alerts. Platform updates add SSL verification controls, automatic deployment cleanup for deleted sources, and improved gateway token security. Gateway 1.25.10 introduces SSL verification control and configurable update windows for streamlined updates.

Nodes

• Generic Detector: Use GPU to export model with FP16, run with CUDA by default and make TRT engine optional to improve start times. Also adds support for parallelized inference engine to reduce memory consumption.
• Filter Objects : remove / retain non specified objects, and add "object within object" filters (#298)
• Publish and Receive MQTT Messages : Explicitly specify MQTT version
• Trigger Google Chat Alert : Support external certificate files.

Platform

• 🍀 Allow turning off SSL cert verification for streams using the disable cert verification flag in stream and camera settings
• Delete a source's deployments when the video source is deleted
• Prevent gateway tokens from accessing routes for other gateways

Gateway

1.25.10 (2024-03-27)

• 🍀 Allow turning off SSL cert verification for streams that have the disable cert verification flag set.
• 🍀 Support for update windows. Set time range during which to check for Gateway updates using the installer script.

New nodes for Roboflow models, camera reporting, face recognition, and Google Chat enhance AI and alerting capabilities. Deployments now inherit labels from cameras and streams. Gateway 1.23.67 introduces multi-GPU load balancing, custom object insertion, improved deployment stop behavior, and prevents runaway restarts for file-based deployments.

Nodes

• 🍀 AI Model by Roboflow : Added support for Roboflow models, expanding integration options for AI-driven workflows
• Publish to Lumeo Dashboard : The Lumeo dashboard node now reports cameras as sources, improving source visibility in workflows.
• Face Recognition Edge : Updated face recognition with new models, yaw calculation, and other improvements for better accuracy.
• Trigger Google Chat Alert : Custom action button support added for Google Chat integration.

Platform

• Deployments now inherit Labels from cameras & streams

Gateway

1.23.67 (2024-02-22)

• Multi-gpu load-balancing support in parallelized inference engine
• Enable custom functions to insert new objects
• Stop restarting file deployments in case of errors to prevent runaway restart loops when the source is corrupted or yields a pipeline error
• Improve deployment stop behavior

New and updated nodes for Axis IO, Jetson GPIO, MQTT, speed estimation, and webhooks enhance integration and functionality. Platform gains security updates, API pagination, and improved gateway performance. Gateway release 1.22.17 fixes deployment and WebRTC issues while adding YOLO ONNX model support.

Nodes

• 🍀 Axis Network IO Read and Axis Network IO Write nodes make it easy to integrate with devices like sirens, strobes or other digital inputs.
• 🍀 Jetson GPIO Read Node and Jetson GPIO Write Node to read and write digital inputs on Jetson devices.
• 🍀 Publish and Receive MQTT Messages lets you integrate with MQTT for IOT device communications
• Speed Estimation node updated to support pps (pixels per second) unit and add motion types (stop_and_go, hover, regular).
• Send HTTP Webhook now lets you send webhooks at the start or end of the pipeline.

Platform

• Security updates and enhancements; support for new single-use Gateway provisioning tokens.
• Improved performance of starting and stopping cloud gateways
• Add pagination to all API endpoints.

Gateway (1.22.17)

1.22.17 (2024-01-29)

• Improved deployment state management and fix for deployments stuck in stopping state
• Support for new Lumeo YOLO ONNX model format in the parallelized inference engine
• Fix WebRTC playback issues

New quick-deploy workflow, GPT4 integration and large vision model nodes!

Platform

• 🍀 Quick Deploy workflow pre-built analytics en-masse to multiple cameras with a few clicks. This workflow is designed for integration and deployment teams to speed up analytics rollouts. Get started here : Quick Deploy Guide

Nodes

• 🍀 OCR to perform text recognition on regions of a frame or specific objects
• 🍀 OpenAI GPT4 Lookup looks up frame regions or objects using OpenAI GPT-4 APIs.
• 🍀 Generic Detector detects objects by providing their description
• 🍀 Generic Classifier classifies objects or frame regions by providing their description
• 🍀 Caption Generator captions a scene

Crowd and NSFW models, Labels for files, Improved streaming and new nodes for interfacing with external systems!

Hot off the press, here are the highlights from the latest release:

Platform

• Updated Model format support
  • 🍀 New format support : Azure YOLO CustomVision (General Compact S2) is now supported.
  • Custom models can now be set to use Parallelized (enhanced) or Default inference engine. Parallelized uses Triton to reduce memory consumption, but some model formats may not be supported. If you run into errors with the Parallelized model, you can fall back to Default.
• Files generated by Save Clip Node and Save Snapshot Node now support Labels and description for manual review workflows. Dashboard Files widget also supports displaying labels and description. Click on a file to view and set.
• Dashboards enhancements :
  • New metrics for data widgets : Latest Queue Size, Latest Estimated Wait time
  • Multistream widget now lets you enable / disable autoplay. When disabled, it periodically refreshes stream snapshots.

Nodes

• 🍀 Jetson GPIO Read Node and Jetson GPIO Write Node let you read and write GPIO pins on Jetson devices to integrate with external systems such as strobes and burglar alarms.
• 🍀 Webhook Receiver Node lets you receive webhooks from external services and inject the data into the pipeline. (requires Gateway 1.22.13).
• 🍀 AXIS Network IO Write Node
• Occupancy Monitor new metadata : estimated wait times, computed using Little's law.
• Trigger Google Chat Alert Node now supports serving media locally. When set, it will start a webserver to serve locally saved media for that specific deployment, and insert a media url that references itself. This is designed to help improve notification latencies for slow networks where uploads can take long time.
• Time Trigger Node now supports multiple, freeform time ranges so you can specify multiple times without having to use multiple time trigger nodes. Ex. weekdays 9am-5pm.
• Trigger Google Chat Alert Node, Trigger SMS Alert Node, Trigger Email Alert Node now support Jinja style templates in the message body fields so you can use metadata to populate the notification body.

Models

• 🍀 High Density Crowd model makes it possible to count hundreds of people in very high density environments such as events and concerts.
• 🍀 NSFW model detects content that can be classified as "Not-safe-for-work".

Gateway (1.22.13)

• Enhanced, persistent logging and diagnostics retrieval.
• Parallelized (Triton) inferencing now supports ONNX models.
• Supports Azure CustomVision Yolo models.
• Support for reading and writing GPIO pins on Jetson devices.
• Adds onnxruntime-gpu that can be used by custom function nodes to run custom models.
• Bug fixes:
  • Auto restart the deployments when GPU is unavailable because of driver updates.
  • Reduced the amount of time it takes to stop deployments.
  • RTSP and WebRTC playback is snappier and more reliable.

Role-based Access Controls, Lumeo-Ready Gateways, Time-bound trial gateways and RTSP playback in Console.

Hot off the press, here are the highlights from the latest release:

Platform

• 🍀 Roles Based Access Control : Assign organization and workspace level roles to limit what users can do. Available on Enterprise plans.
• 🍀 Lumeo-Ready Gateways : Lumeo-Ready gateways are Jetson and DGPU based devices that ship with Lumeo pre-installed, and make setup easy via a web interface in the field. You can order Lumeo-Ready gateways from select partners.
• 🍀 Lumeo Console now allows playback of RTSP streams from Cameras and pipeline output streams. Support to playback user-configured RTSP streams is coming next. (Requires Gateway v 1.19.1 or higher).
• Time-bound Trial Gateways : We heard you! Anyone can now create a Lumeo Cloud Trial Gateway with a 3 hour runtime. After 3 hours, the gateway will automatically shut down, and you can restart it in the Console. This allows us to offer Trial gateways to more users and never have to expire them.

Gateway (1.19.15)

• Supports Lumeo Console playback of RTSP streams.
• Triton-based (parallelized) inferencing now supports Nvidia ETLT models.
• Reduced memory utilization by lazy loading CUDA
• Supports exporting Metrics using OTLP HTTP push in addition to previously supported OTLP gRPC push.

Dashboard enhancements, auto-triggers to save you from writing triggers and Triton-based inferencing increases channel capacity by 2-3x!

Hot off the press, here are the highlights from the latest release:

Platform

• Dashboards now support Heatmaps, Grouped Bar charts and Stacked Bar charts.

Nodes

• 🍀 Webhook Node supports posting object snapshots.
• 🍀 Trigger Google Chat Alert Node now sends media delayed as a threaded attachment so that the initial alert goes out as soon as possible.
• 🍀 24x7 Incident Management System node to create an incident in 24x7 IMS
• 🍀 Save GIF Node : Saves a GIF based on trigger conditions.
• 🍀 Auto Triggers for Webhook Node, Publish to Google Sheets, Publish to Elasticsearch Node : No need to specify trigger conditions explicitly. Select Auto trigger mode to automatically publish metadata whenever significant data changes, at the specified interval.

  🚧

  Update existing nodes

  ⚠️ If you used these nodes previously, you should update your pipeline and choose "Custom Trigger" as the trigger type to preserve your existing triggers.

Gateway (1.16.25)

• Support for Triton-based inferencing that reduces memory consumption, allowing more pipelines to be run concurrently.
• Performance improvements improve inferencing and throughput by 1.5x
• Improved debug logs make it easier to find and fix issues.

Multimodal Data, Video and Image Dashboards, Reidentification Tracker and Models!

Hot off the press, here are the highlights from the latest release:

Platform

• Dashboards : Introducing multi-modal dashboards that make it easy to visualize live streams, snapshots, gifs and data extracted from video.

Nodes

• 🍀 Smart Zoom
• 🍀 ReIdentify Objects node
• 🍀 Trigger Google Chat Alert Node
• 🍀 Publish to Lumeo Dashboard node makes it easy to publish metadata from any Rule nodes to the Lumeo dashboard.

Models

• Person Reidentification Model
• Vehicle Reidentification Model

Gateway (1.15.10)

• RTSP server now listens on all NICs for output RTSP streams
• New Tracker type : "Re-Id". Uses DeepSORT tracking algorithm to perform tracking using AI models for similarity matching.
• Balena: Support to store models and configuration on an external partition using LUMEO_CONFIG_PARTITION env variable set in Balena device configuration.

Universal Bridge, Labels, Solink and Rhombus integrations, 360 People models, OSHA compliance and a host of other goodies packed into a fun action-packed spring release!

Hot off the press, here are the highlights from the latest release:

Platform

• 🍀 Universal Bridge : New cloud-based service that lets you add AI to any IP camera by configuring it to send clips to Lumeo's cloud on motion or at periodic intervals. Process these clips using any custom analytic designed in the Lumeo cloud. Read our press release here.
• 🍀 Labels : Simplify management of Cameras, Gateways, Streams using labels. Labels are hierarchical (nest them all you like), and provide a filtered view of your deployed infrastructure.

• A brand new Console UX that makes it easier to navigate and get common tasks done!

Nodes

Integrations

• 🍀 Trigger Solink Event Node : Send Events to Solink when an event occurs in Lumeo. Read more about the Solink integration here
• 🍀 Trigger Rhombus Event : Send Events to
  Rhombus
  when an event occurs in Lumeo.

Recognitions & Rules

• 🍀 Scene Change Detection : Detect scene changes in the camera view. Useful for OSHA compliance (Box stacking, pathways and fire extinguishers unblocked, and more).
• 🍀 Aggregate Metadata Node : Aggregate metadata with Wildcards and output a single trigger. Simplify your Trigger Conditions.
• 🍀 Merge Objects Node : Combine objects from multiple models (such as multiple people detection models) when they overlap. Makes it easy to build pipelines using different models such as day and night people models.
• 🍀 Time Trigger Node : Trigger at certain times of the day, periodic intervals, etc.
• Face Recognition Edge Node : Speed up face loading for edge face rec by 10x
• Webhook Node : New static key format for Webhooks, also image and snapshot POSTS

Models

• 🍀 Person and Vehicle Reidentification Models
• 🍀 360 (Fisheye) camera people detection Models (Premium)
• 🍀 2D Bodypose estimation model

Gateway (v 1.10.10)

• Support for Nvidia Orin devices
• Use your own S3 buckets to store clips and snapshots. Configure in the Save Clip Node and Save Snapshot Node.
• Collect and report runtime metrics to an OpenTelemetry compatible receiver. Configure under Settings -> Metrics
• Bug fixes and performance updates.

Platform

• 🍀 GCP - Kubernetes Guide
• 🍀 Deployment limits per gateway : limit how many deployments can run in parallel on a single gateway
• 🍀 Console UX enhancements: enhanced JSON log viewer, Snapshot previews in streams, incremental loading for large number of cameras/deployments/resources.

Nodes

• 🍀 Webhook Node updates to push snapshots and clips with Form multi-part POST.
• 🍀 Camera calibration Node support
• 🍀 Occupancy Monitor updated to add exited_objects and objected_entered_timestamp to metadata.
• 🍀 Face Recognition Edge Node for Face recognition via the edge (local) processing.
• 🍀 Blur Objects Node now supports conditional blurring

Gateway (v 1.6.22)

• 🍀 Deepstream 6.1 support, Jetson requires Jetpack 5.0, DGPU requires drivers 515.x+, Balena will work as is
• 🍀 gateway_id and gateway_name, pipeline_id, pipeline_name available in function node parameters
• 🍀 Set "Tensor::source_node_id" and add "source" to object and object attribute meta so you can easily tell which node/model generated that object type and attributes.
• 🍀 Support for processing models trained using Azure Custom Vision Service (https://customvision.ai)
• 🍀 [Balena] Auto assign gateway name from BALENA_DEVICE_NAME_AT_INIT env var if present
• 🍀 Add extra metadata to video source node
• 🐛 Reduce MQTT traffic when there are no subscribers to logs topic
• 🐛 On Jetson and dGPU, local snapshot nodes save every frame. This has been fixed.