Frigate NVR Skill

Complete guide for managing Frigate Network Video Recorder with AI object detection.

Quick Reference

Core Components

Component	Purpose
Frigate Server	Main NVR processing and recording
Detector	AI inference (Coral TPU, OpenVINO, CPU)
go2rtc	RTSP/WebRTC streaming
MQTT	Event publishing to Home Assistant

Configuration File Location

/config/frigate.yml  # Main configuration
/media/frigate/      # Recording storage

---

1. Installation

Docker Compose

version: "3.9"
services:
  frigate:
    container_name: frigate
    image: ghcr.io/blakeblackshear/frigate:stable
    restart: unless-stopped
    privileged: true
    shm_size: "256mb"
    devices:
      - /dev/bus/usb:/dev/bus/usb  # Coral USB
      # - /dev/apex_0:/dev/apex_0  # Coral PCIe
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - ./config:/config
      - ./media:/media/frigate
    ports:
      - "5000:5000"   # Web UI
      - "8554:8554"   # RTSP
      - "8555:8555"   # WebRTC
    environment:
      FRIGATE_RTSP_PASSWORD: "password"

With Coral TPU

devices:
  - /dev/bus/usb:/dev/bus/usb  # USB Coral
  # OR for PCIe Coral:
  - /dev/apex_0:/dev/apex_0

---

2. Configuration

Basic Camera Setup

# frigate.yml
mqtt:
  enabled: true
  host: mqtt-broker
  port: 1883
  user: frigate
  password: "{FRIGATE_MQTT_PASSWORD}"

detectors:
  coral:
    type: edgetpu
    device: usb

cameras:
  front_door:
    ffmpeg:
      inputs:
        - path: rtsp://user:pass@192.168.1.100:554/stream1
          roles:
            - detect
            - record
    detect:
      width: 1280
      height: 720
      fps: 5
    objects:
      track:
        - person
        - car
        - dog
        - cat
    record:
      enabled: true
      retain:
        days: 7
        mode: motion
    snapshots:
      enabled: true
      retain:
        default: 14

Detection Zones

cameras:
  front_door:
    zones:
      driveway:
        coordinates: 0,500,400,500,400,0,0,0
        objects:
          - car
        filters:
          car:
            min_area: 5000
      porch:
        coordinates: 400,500,640,500,640,0,400,0
        objects:
          - person
          - package

Motion Masks (Ignore Areas)

cameras:
  front_door:
    motion:
      mask:
        - 0,0,200,0,200,100,0,100  # Top-left corner

Object Filters

cameras:
  front_door:
    objects:
      track:
        - person
        - car
      filters:
        person:
          min_area: 1000
          max_area: 100000
          min_score: 0.6
          threshold: 0.7
        car:
          min_area: 5000
          min_score: 0.5

---

3. Recording Configuration

Recording Modes

record:
  enabled: true
  retain:
    days: 7           # Keep all recordings for 7 days
    mode: motion      # all, motion, active_objects
  events:
    pre_capture: 5    # Seconds before event
    post_capture: 5   # Seconds after event
    retain:
      default: 30     # Keep events for 30 days
      mode: motion

Storage Management

record:
  retain:
    days: 3
    mode: motion
  events:
    retain:
      default: 14
      objects:
        person: 30    # Keep person events longer
        car: 7

---

4. Detector Configuration

Google Coral USB

detectors:
  coral:
    type: edgetpu
    device: usb

Google Coral PCIe

detectors:
  coral:
    type: edgetpu
    device: pci

Multiple Corals

detectors:
  coral1:
    type: edgetpu
    device: usb:0
  coral2:
    type: edgetpu
    device: usb:1

OpenVINO (Intel CPU/GPU)

detectors:
  ov:
    type: openvino
    device: AUTO
    model:
      path: /openvino-model/ssdlite_mobilenet_v2.xml

CPU (Fallback)

detectors:
  cpu:
    type: cpu
    num_threads: 4

---

5. go2rtc Streaming

Stream Configuration

go2rtc:
  streams:
    front_door:
      - rtsp://user:pass@192.168.1.100:554/stream1
    front_door_sub:
      - rtsp://user:pass@192.168.1.100:554/stream2

cameras:
  front_door:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/front_door
          input_args: preset-rtsp-restream
          roles:
            - record
        - path: rtsp://127.0.0.1:8554/front_door_sub
          input_args: preset-rtsp-restream
          roles:
            - detect

WebRTC for Low Latency

go2rtc:
  webrtc:
    candidates:
      - 192.168.1.50:8555  # Your Frigate IP
      - stun:8555

---

6. Home Assistant Integration

MQTT Configuration

mqtt:
  enabled: true
  host: 192.168.1.10
  port: 1883
  topic_prefix: frigate
  client_id: frigate
  user: mqtt_user
  password: mqtt_pass

Home Assistant Config

# configuration.yaml
mqtt:
  sensor:
    - name: "Frigate Front Door Person"
      state_topic: "frigate/front_door/person"

camera:
  - platform: mqtt
    name: "Front Door"
    topic: "frigate/front_door/person/snapshot"

Frigate Integration (HACS)

# Use the Frigate integration from HACS for:
# - Camera entities
# - Binary sensors for objects
# - Event notifications
# - Lovelace card

---

7. Notifications

MQTT Events

# Events published to:
frigate/events                    # All events
frigate/<camera>/person          # Person count
frigate/<camera>/person/snapshot # Snapshot image

Event JSON Structure

{
  "type": "new",
  "before": {},
  "after": {
    "id": "1234567890.123456-abc123",
    "camera": "front_door",
    "frame_time": 1234567890.123,
    "label": "person",
    "score": 0.87,
    "box": [100, 200, 300, 400],
    "area": 20000,
    "region": [0, 0, 640, 480],
    "current_zones": ["driveway"],
    "entered_zones": ["driveway"]
  }
}

---

8. API Reference

REST API Endpoints

# Get config
curl http://frigate:5000/api/config

# Get camera snapshot
curl http://frigate:5000/api/front_door/latest.jpg -o snapshot.jpg

# Get events
curl "http://frigate:5000/api/events?camera=front_door&limit=10"

# Delete event
curl -X DELETE http://frigate:5000/api/events/EVENT_ID

# Restart Frigate
curl -X POST http://frigate:5000/api/restart

# Get recording summary
curl http://frigate:5000/api/front_door/recordings/summary

MQTT Commands

# Enable/disable detection
mosquitto_pub -t "frigate/front_door/detect/set" -m "ON"
mosquitto_pub -t "frigate/front_door/detect/set" -m "OFF"

# Enable/disable recording
mosquitto_pub -t "frigate/front_door/recordings/set" -m "ON"

# Enable/disable snapshots
mosquitto_pub -t "frigate/front_door/snapshots/set" -m "ON"

---

9. Performance Tuning

Reduce CPU Usage

cameras:
  front_door:
    detect:
      fps: 5          # Lower FPS for detection
      width: 1280     # Lower resolution
      height: 720
    ffmpeg:
      output_args:
        detect: -f rawvideo -pix_fmt yuv420p
        record: preset-record-generic-audio-aac

Memory Optimization

# Docker Compose
shm_size: "256mb"  # Increase if needed

# frigate.yml
cameras:
  front_door:
    ffmpeg:
      hwaccel_args: preset-vaapi  # Intel Quick Sync
      # hwaccel_args: preset-nvidia  # NVIDIA GPU

Hardware Acceleration

ffmpeg:
  hwaccel_args: preset-vaapi  # Intel
  # hwaccel_args: preset-nvidia-h264  # NVIDIA
  # hwaccel_args: preset-rpi-64-h264  # Raspberry Pi 4

---

10. Troubleshooting

Common Issues

Camera not connecting:

# Test RTSP stream
ffprobe rtsp://user:pass@192.168.1.100:554/stream1

# Check Frigate logs
docker logs frigate 2>&1 | grep "front_door"

High CPU usage:

• Lower detect FPS to 5
• Use hardware acceleration
• Add motion masks for busy areas
• Use sub-stream for detection

Coral not detected:

# Check USB devices
lsusb | grep Google

# Check Frigate logs
docker logs frigate 2>&1 | grep -i coral

No recordings:

# Check storage permissions
ls -la /media/frigate/

# Check recording config
grep -A 10 "record:" /config/frigate.yml

Debug Configuration

logger:
  default: info
  logs:
    frigate.record: debug
    frigate.event: debug
    detector.coral: debug

---

Best Practices

1. Use sub-streams for detection (lower resolution, less CPU)
2. Use main stream for recording (full quality)
3. Set appropriate FPS - 5 FPS is usually sufficient for detection
4. Configure zones to reduce false positives
5. Use motion masks for trees, roads, reflections
6. Regular maintenance - prune old recordings
7. Monitor storage - set appropriate retention policies
8. Use Coral TPU for best detection performance