Observatory Pipeline Helm Chart

This chart deploys Observatory Pipelines to a Kubernetes cluster.

The implementation of the Helm chart is right now the bare minimum to get it to work.

Usage in Teknoir platform

Use the HelmChart to deploy the Observatory Pipeline to a Device.

---
apiVersion: helm.cattle.io/v1
kind: HelmChart
metadata:
  name: observatory-pipeline
  namespace: default
spec:
  repo: https://teknoir.github.io/observatory-pipeline-helm
  chart: observatory-pipeline
  targetNamespace: default
  valuesContent: |-
    # Example for minimal configuration
    streams:
      - stream: update-frame-rate-control
    
    # Values possible for Observatory Pipelines
    defaults:
      stream: update-frame-rate-control
      image:
        repository: us-docker.pkg.dev/teknoir/gcr.io/observatory-event-processing
        tag: dev-4552014
      # Set the log level for the nvr
      # - DEBUG
      # - INFO
      # - WARNING
      # - ERROR
      debugLevel: INFO
      mqtt:
        # Local MQTT broker
        broker: mqtt.kube-system
        port: 1883
        username: "emqx_operator_controller"
        password: ""
        qos: 2
        keepAliveInterval: 10
        connectRetryInterval: 10
        # Topics
        topicSink: "events/sink"
        topicSource: "events/source"
        topicAllCameras: "camera/frames/#"
        topicDetections: "events/detections"
        topicInference: "events/inference"
        topicMovements: "events/movements"
        topicFilteredMovements: "events/filtered-movements"
        topicAlerts: "events/alerts"
        topicStaleAlerts: "events/stale-alerts"
        topicSensors: "events/sensors"
        topicMeasurements: "events/measurements"
        topicToeEvents: "toe/events"
        topicToeCommands: "toe/commands/devstudio"
        topicAllDevices: "/devices/+/events"
      resources:
        limits:
          cpu: 2000m
          memory: 2048Mi
        requests:
          cpu: 100m
          memory: 128Mi
      fe:
        frontEndDefaultInterval: 60 # seconds
      historian:
        apiUrl: "http://historian-api:8000"
        recordDetectionsInterval: 333 # milliseconds
        objectsToFilter: "car,truck,motorbike,person,face" # supports multiple labels via comma separated list
      regions:
        iouThreshold: 0.001 # default Intersection Over Union (IOU) threshold for object detection inside region
      video:
        minStartBeforeAlert: 10 # seconds
        mountSegmentPath: false
        segmentPath: "/opt/teknoir/video/segments"
        mountClipTempPath: false
        clipTempPath: "/opt/teknoir/video/clips"
        segmentLength: 15 # seconds
        segmentRetentionLength: 5760 # segments
        maxClipLength: 600 # seconds
        clipRunupLength: 10 # seconds
        chunkMaxSize: 500000 # bytes
        chunkRetentionSeconds: 300 # seconds
        videoBasePath: "media/videos"
        hasAnnotations: true
        annotationsBucketPath: "media/annotations"
        snapshotBucketPath: "media/snapshots"
        staleSeconds: 30 # seconds
      alerts:
        personDownInterval: 10
        personDownScoreThreshold: 0.85
        personUpScoreThreshold: 0.75
        personDownUpTotalCountThreshold: 10
        loiteringLabels: "person" # supports multiple labels via comma separated list
        sensorsNodeData: "{}"
        loiteringThreshold: 60 # 60 seconds
        loiteringCheckStationary: true
        loiteringStationaryThreshold: 0.01 # 1% of FOV
        repetitiveAlertsCheck: true
        repetitiveAlertsTimespan: 3600 # seconds
        repetitiveAlertsIouThreshold: 0.9 # 90%
        safeOpenWindowSizeSeconds: 5 # Default window size of 5 seconds
        numberOfDominantLabels: 5
        motionAlertMinInterval: 30 # 30 seconds

Example values.yaml

defaults:
  debugLevel: DEBUG

streams:
  - stream: update-frame-rate-control
    mqtt:
      topicDetections: "camera/frames/#"
  - stream: record-detections
    mqtt:
      topicDetections: "camera/frames/#"
  - stream: update-movements
  - stream: filter-movements
  # ALERTS
  - stream: motion
  - stream: loitering
  - stream: person-down
  - stream: catalytic-theft
  # ALERT PROCESSING
  - stream: update-alerts
  - stream: make-clip
    video:
      mountSegmentPath: true
      mountClipTempPath: true

Adding the repository

helm repo add teknoir-observatory-pipeline https://teknoir.github.io/observatory-pipeline-helm/

Installing the chart

helm install observatory-pipeline teknoir-observatory-pipeline/observatory-pipeline -f values.yaml