微软人脸识别与认证

操作步骤

1. 准备标准照待用
2. 在微软人脸云服务中配置组、用户、标准照片
3. 微软人脸识别配置解说
4. 使用vlc配置一个抓屏摄像头（同上一个视频）
5. 效果实验

参考

• 微软人脸云服务API说明

https://westcentralus.dev.cognitive.microsoft.com/docs/services/563879b61984550e40cbbe8d/operations/563879b61984550f30395236

• 微软人脸云端数据结构

• 相关API调用命令

• 创建组

  curl -X PUT "https://westcentralus.api.cognitive.microsoft.com/face/v1.0/persongroups/presidents" \

  -H "Content-Type: application/json" \

  -H "Ocp-Apim-Subscription-Key: 4211c17af9b14ff581a41492bd7b069b" \

  --data-ascii "{'name': 'PresidentsInUS', 'userData': 'Presidents of United States'}"

• 创建用户

  curl -X POST "https://westcentralus.api.cognitive.microsoft.com/face/v1.0/persongroups/presidents/persons" \

  -H "Content-Type: application/json" \

  -H "Ocp-Apim-Subscription-Key: 4211c17af9b14ff581a41492bd7b069b" \

  --data-ascii "{'name': 'Clinton', 'userData': 'Bill Clinton'}"

• 上传用户照片

  curl -X POST \

  "https://westcentralus.api.cognitive.microsoft.com/face/v1.0/persongroups/presidents/persons/bb6d222a-8956-40ee-8a78-e00ccd3c6503/persistedFaces" \

  -H "Ocp-Apim-Subscription-Key: 4211c17af9b14ff581a41492bd7b069b" \

  -H "Content-Type: application/octet-stream" \

  --data-binary "@/home/pi/Pictures/clinton.jpg"

• 训练组

  curl -X POST "https://westcentralus.api.cognitive.microsoft.com/face/v1.0/persongroups/presidents/train" \

  -H "Ocp-Apim-Subscription-Key: 4211c17af9b14ff581a41492bd7b069b" \

  --data-ascii ""

• HA中微软人脸识别配置说明文档

https://www.home-assistant.io/integrations/microsoft_face_identify

• 微软人脸识别与认证配置样例（example_9_5_1.yaml）

  example_9_5_1.yaml

  microsoft_face: api_key: 4211c17af9b14ff581a41492bd7b069b azure_region: westcentralus

  image_processing: - platform: microsoft_face_identify scan_interval: 1000000 group: presidents confidence: 10 source: - entity_id: camera.cam_input name: ms_face_identify

  script: ms_face_identify_script: alias: 微软人脸识别 sequence: - service: image_processing.scan data: entity_id: image_processing.ms_face_identify

  automation: - alias: Clinton Identify trigger: platform: event event_type: image_processing.detect_face event_data: entity_id: image_processing.ms_face_identify name: 'Clinton' action: service: persistent_notification.create data_template: title: '人脸认证' message: '克林顿出现在摄像头中，可信度{{ trigger.event.data.confidence }}%'

• VLC http mjpeg输出配置（同上一篇）

:sout=#transcode{vcodec=MJPG,vb=800,scale=自动,acodec=none,scodec=none}:standard{access=http{mime=multipart/x-mixed-replace; boundary=7b3cc56e5f51db803f790dad720ed50a},mux=mpjpeg,dst=:8888/} :no-sout-all :sout-keep

返回课程目录↩