seeed-projects/seeed-jetson-developtool/deploy-frigate

Deploy Frigate on Jetson

Frigate is an open-source NVR with real-time AI object detection. This skill deploys it on Jetson via Docker with TensorRT-accelerated YOLOv7 for IP camera surveillance.

Execution model

Run one phase at a time. After each phase:

• Relay all output to the user.
• If output contains [STOP] → stop immediately, consult the failure decision tree.
• If output ends with [OK] → tell the user "Phase N complete" and proceed.

Prerequisites

Phase 1 — Install Docker and dependencies (~5 min)

sudo apt-get update && sudo apt-get upgrade -y
sudo apt-get install -y python3-pip python3-dev python3-venv \
    build-essential libssl-dev libffi-dev git \
    apt-transport-https ca-certificates curl software-properties-common

Install Docker:

curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
sudo add-apt-repository "deb [arch=arm64] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable"
sudo apt-get update
sudo apt-get install -y docker-ce

Install Docker Compose:

sudo pip3 install docker-compose
docker-compose --version

Install NVIDIA Container Toolkit per https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html

[OK] when docker-compose --version prints a version. [STOP] if Docker install fails.

Phase 2 — Pull Frigate Docker image (~5–10 min)

docker pull ghcr.io/blakeblackshear/frigate:stable-tensorrt-jp5

Use stable-tensorrt-jp4 for JetPack 4.6 devices.

[OK] when pull completes. [STOP] if pull fails — check network and Docker permissions.

Phase 3 — Create Frigate configuration (~5 min)

mkdir -p ~/frigate/config ~/frigate/storage

Create ~/frigate/config/config.yml with your camera RTSP URL:

mqtt:
  enabled: False
 
cameras:
  my_camera:
    enabled: True
    ffmpeg:
      hwaccel_args: preset-jetson-h264
      inputs:
        - path: rtsp://admin:password@<CAMERA_IP>:554/cam/realmonitor?channel=1&subtype=0
          roles:
            - detect
 
birdseye:
  enabled: True
  mode: objects
 
detectors:
  tensorrt:
    type: tensorrt
    device: 0
 
model:
  path: /config/model_cache/tensorrt/yolov7-320.trt
  input_tensor: nchw
  input_pixel_format: rgb
  width: 320
  height: 320
 
detect:
  fps: 20
  width: 1280
  height: 720
 
objects:
  track:
    - person

Create ~/frigate/docker-compose.yml:

services:
  frigate:
    privileged: true
    environment:
      - YOLO_MODELS=yolov7-320
      - USE_FP16=false
    container_name: frigate
    runtime: nvidia
    volumes:
      - /home/${USER}/frigate/config:/config
      - /home/${USER}/frigate/storage:/media/frigate
    ports:
      - "5000:5000"
      - "8554:8554"
    image: ghcr.io/blakeblackshear/frigate:stable-tensorrt-jp5

[OK] when both files are created.

Phase 4 — Deploy and verify Frigate (~3 min)

cd ~/frigate
docker-compose up -d
docker ps | grep frigate

Check logs for TensorRT initialization:

docker logs frigate

[OK] when container is running and logs show TensorRT model loaded. [STOP] if container fails to start.

Phase 5 — Access Frigate web interface

Open a browser on the same network and navigate to:

http://<jetson-ip>:5000

You should see the Frigate dashboard with live camera feeds and detection events. Detection latency should be ~33ms per frame on Xavier NX.

[OK] when the dashboard loads with live video.

Failure decision tree

Reference files

• references/source.body.md — Full Seeed Wiki tutorial with hardware prep, detailed config explanations, and troubleshooting (reference only)