玩了下CompreFace

官网
github

本地测试了一下几种不同算法的时间

如果用Mobilenet

{'result': [{'box': {'probability': 0.8101944327354431, 'x_max': 114, 'y_max': 573, 'x_min': 39, 'y_min': 480}, 'mask': {'probability': 0.9885141253471375, 'value': 'without_mask'}, 'execution_time': {'detector': 33.0, 'mask': 31.0}}], 'plugins_versions': {'detector': 'insightface.FaceDetector@retinaface_mnet025_v1', 'mask': 'facemask.MaskDetector'}}

速度很快，但是人脸检测，检出率很低

FaceNet

{'result': [{'box': {'probability': 0.9974731802940369, 'x_max': 117, 'y_max': 585, 'x_min': 30, 'y_min': 477}, 'mask': {'probability': 1.0, 'value': 'without_mask'}, 'execution_time': {'detector': 315.0, 'mask': 65.0}}], 'plugins_versions': {'detector': 'facenet.FaceDetector', 'mask': 'facemask.MaskDetector'}}

SubCenter-ArcFace-r100

{'result': [{'box': {'probability': 0.9747834205627441, 'x_max': 102, 'y_max': 579, 'x_min': 36, 'y_min': 489}, 'mask': {'probability': 0.9936507344245911, 'value': 'without_mask'}, 'execution_time': {'detector': 315.0, 'mask': 45.0}}], 'plugins_versions': {'detector': 'insightface.FaceDetector@retinaface_r50_v1', 'mask': 'facemask.MaskDetector'}}

速度也不快，检出率也不高

默认配置

{'result': [{'box': {'probability': 0.9974731802940369, 'x_max': 117, 'y_max': 585, 'x_min': 30, 'y_min': 477}, 'mask': {'probability': 1.0, 'value': 'without_mask'}, 'execution_time': {'detector': 488.0, 'mask': 56.0}}], 'plugins_versions': {'detector': 'facenet.FaceDetector', 'mask': 'facemask.MaskDetector'}}

人脸检出率很高

关于GPU，官方文档在linux上没什么问题，我的环境是笔记本windows11+docker desktop+3050
按照官方的

  compreface-core:
    image: ${registry}compreface-core:${CORE_VERSION}
    restart: always
    container_name: "compreface-core"
    runtime: nvidia
    environment:
      - ML_PORT=3000
      - UWSGI_PROCESSES=${uwsgi_processes:-1}
      - UWSGI_THREADS=${uwsgi_threads:-1}

这个runtime: nvidia这个始终配置不上去，后来发现这样就可以了

  compreface-core:
    image: ${registry}compreface-core:${CORE_VERSION}
    restart: always
    container_name: "compreface-core"
    environment:
      - ML_PORT=3000
      - UWSGI_PROCESSES=${uwsgi_processes:-2}
      - UWSGI_THREADS=${uwsgi_threads:-1}
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]