• ./ice.exe --client --cmd 1 --id 2 --model-conf-path http://192.169.4.113:8084/ --detector-conf @--detect-model@models/Face_BITMAINLAND/DETECT.conf@--feature-model@models/Face_BITMAINLAND/FEATURE.conf@ --comparison-list http://192.169.4.113:8084/platform/images/recognize/研发中心/20221214211140.txt

  • ./ice.exe --client --cmd 1 --id 2 --model-conf-path http://192.169.4.113:8084/ --detector-conf @--detect-model@models/Face_NVIDIA_1080/DETECT.conf@--feature-model@models/Face_NVIDIA_1080/FEATURE.conf@ --comparison-list http://192.169.4.113:8084/platform/images/recognize/研发中心/20221214211140.txt

  • ./ice.exe --client --cmd 1 --id 2 --model-conf-path http://192.169.4.113:8084/ --detector-conf @--detect-model@models/Face_BITMAINLAND/DETECT.conf@--feature-model@models/Face_BITMAINLAND/FEATURE.conf@ --comparison-list http://192.169.4.113:8084/platform/images/recognize/研发中心/20221214211140.txt

人脸识别

  • FEATURE.conf文件中的参数: --feature-metric必须设置为1，也就是设定输出的分数为相似度

  • ice.exe的参数--feature-metric默认为1，将模型输出的数值，乘以100

  • 模型，必须先加载人脸检测模型DETECT.conf，然后是特征提取模型FEATURE.conf

bitmainland

  • ./ice.exe --client --server-order-protocol 1 --server-zmq-address tcp://127.0.0.1:9100 --cmd 3 --id 12 --parse-file ./base64-file.txt

  • sudo docker run -itd --restart=always --privileged=true --name face -p 8083:8083 -p 8084:8084 -p 16379:6379 -v /system:/system -v /etc/localtime:/etc/localtime:ro -v /etc/timezone:/etc/timezone:ro -e LOCAL_USER_ID=id -u -e PYTHONPATH=/system/lib -e LANG=C.UTF-8 -e LC_ALL=C.UTF-8 mjpeg:face /bin/bash /etc/init.d/start_server.sh