Running Frigate NVR on Raspberry Pi for Security Cameras

Frigate is an open-source NVR with AI object detection. Run it on a Pi 4/5 for a local, private security camera system.

Why Frigate

• 100% local — no cloud, no subscription
• AI detection (person, car, animal) on every camera
• Integrates with Home Assistant
• Stores recordings locally

Requirements

• Raspberry Pi 4 or 5 (4GB minimum)
• RTSP-capable IP cameras
• USB SSD for recordings (SD card too slow/small)
• Optional: Coral USB TPU for accelerated detection

Install Docker

curl -fsSL https://get.docker.com | sh
sudo usermod -aG docker pi

Frigate Config

Create /home/pi/frigate/config.yml:

mqtt:
  host: 192.168.1.x  # your MQTT broker (or use Home Assistant add-on)

cameras:
  front_door:
    ffmpeg:
      inputs:
        - path: rtsp://admin:[email protected]:554/stream
          roles:
            - detect
            - record
    detect:
      width: 1280
      height: 720
    record:
      enabled: true
      events:
        retain:
          default: 10  # days

Run Frigate

docker run -d --name frigate \
  --restart unless-stopped \
  --privileged \
  -v /home/pi/frigate/config.yml:/config/config.yml \
  -v /mnt/ssd/frigate:/media/frigate \
  -p 5000:5000 -p 1935:1935 \
  ghcr.io/blakeblackshear/frigate:stable

Access web UI at http://raspberrypi.local:5000

Coral USB TPU (Optional)

Plugging in a Coral USB Accelerator speeds up detection from ~200ms to ~10ms per frame — makes real-time detection practical for multiple cameras.