frigate+double-take+codeprojectai 实现人脸识别

对于 frigate 的部署，可以参考前面的文章
codeprojectai 也可以用 deepstack 替代（后者停更好久了）
目前不上云的话，没有比较好的支持 Coral TPU 的离线人脸识别模型方案（或者是我没找到）

codeprojectai 虽然支持 Coral TPU，但是人脸识别模块目前暂不支持

介于上述，这套方案目前物体识别依靠 Coral TPU 可以 10+ms 出结果，但是人脸识别依旧只能依靠 NAS 上可怜的 CPU，几乎需要 1~3s 出结果~对于识别陌生人勉强算是凑合用吧~

docker-compose 文件

services:
  frigate:
    restart: unless-stopped
    container_name: frigate
    image: ghcr.io/blakeblackshear/frigate:stable
    shm_size: 256m # update for your cameras based on calculation above
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - /share/Container/dockerdata/frigate/config:/config
      - /share/MEDIA2/frigate_storage:/media/frigate
      - type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
        target: /tmp/cache
        tmpfs:
          size: 1000000000
#    ports:
#      - "8971:8971" # Authenticated UI and API access without TLS. Reverse proxies should use this port.
#      - "5500:5000"
#      - "8554:8554" # RTSP feeds
#      - "8555:8555/tcp" # WebRTC over tcp
#      - "8555:8555/udp" # WebRTC over udp
    environment:
      - TZ=Asia/Shanghai
      - FRIGATE_RTSP_PASSWORD=frigate_pwd
      - LIBVA_DRIVER_NAME=i965
#    privileged: true
    devices:
      - /dev/bus/usb:/dev/bus/usb
      - /dev/dri/renderD128:/dev/dri/renderD128 # For intel hwaccel, needs to be updated for your hardware
    networks:
      - mynet


  double-take:
    container_name: double-take
    image: skrashevich/double-take
    restart: unless-stopped
    volumes:
      - /share/Container/dockerdata/double-take:/.storage
    environment:
      - TZ=Asia/Shanghai
#    ports:
#      - 3000:3000
    networks:
      - mynet


#  deepstack:
#    container_name: deepstack
#    image: deepquestai/deepstack
#    restart: unless-stopped
#    environment:
#      - TZ=Asia/Shanghai
#      - VISION-FACE=True
#    volumes:
#      - /share/Container/dockerdata/deepstack:/datastore
##    ports:
##      - 5000:5000
#    networks:
#      - mynet


  codeprojectai:
    container_name: codeprojectai
    image: codeproject/ai-server
    restart: unless-stopped
    environment:
      - TZ=Asia/Shanghai
    volumes:
      - /share/Container/dockerdata/codeprojectai/data:/etc/codeproject/ai
      - /share/Container/dockerdata/codeprojectai/modules:/app/modules
#    ports:
#      - 32168:32168
#    devices:
#      - /dev/bus/usb:/dev/bus/usb
    networks:
      - mynet


networks:
  mynet:
    external: true

Coral TPU 直接给了 frigate，如果后续 codeprojectai 人脸识别也支持 TPU 了，可以把 Coral TPU 给它统一处理所有数据！

配置文件

double-take

config.yml

# Double Take
# Learn more at https://github.com/skrashevich/double-take/#configuration
auth: true

mqtt:
  host: 192.168.12.200
  username: mosquitto
  password: mqtt_pwd
# 这里的事件传递给 homeassistant 为后面的消息通知做准备
  topics:
    # mqtt topic for frigate message subscription
    frigate: frigate/events
    #  mqtt topic for home assistant discovery subscription
    homeassistant: homeassistant
    # mqtt topic where matches are published by name
    matches: double-take/matches
    # mqtt topic where matches are published by camera name
    cameras: double-take/cameras


frigate:
# https://frigate-noauth.localnas.top
  url: http://frigate:5000
# 需要获取的 frigate 标签
  labels:
    - person
# 需要获取的摄像头名
  cameras:
    - door

detectors:
#  deepstack:
#    url: http://deepstack:5000
##    key:
#    # number of seconds before the request times out and is aborted
#    timeout: 15
#    # require opencv to find a face before processing with detector
#    opencv_face_required: false
#    # only process images from specific cameras, if omitted then all cameras will be processed
#    cameras:
#     - door

  aiserver:
    url: http://codeprojectai:32168
    # number of seconds before the request times out and is aborted
    timeout: 15
    # require opencv to find a face before processing with detector
    opencv_face_required: false
    cameras:
     - door

time:
  # defaults to iso 8601 format with support for token-based formatting
  # https://github.com/moment/luxon/blob/master/docs/formatting.md#table-of-tokens
  format: yyyy-MM-dd hh:mm:ss
  # time zone used in logs
  timezone: Asia/Shanghai