人脸识别模型试用

前几天用社保APP给我妈进行人脸识别，总是识别失败，今天实验了一个人脸识别模型，本模型可以检测输入图片中人脸以及对应关键点的位置，实现人脸检测和人脸关键点定位二合一。
web方式体验，可以上传照片，然后输出五官位置,从json输出看某个关键点的值为128.32200622558594，其精度已经很高。

json格式输出信息如下：

{
  "detail": "",
  "code": 0,
  "computation_time": "0.12s",
  "data": {
    "boxes": [
      [
        71.2724380493164,
        115.34973907470703,
        288.6782531738281,
        383.5452575683594
      ]
    ],
    "keypoints": [
      [
        128.32200622558594,
        210.29916381835938,
        235.47373962402344,
        209.12826538085938,
        184.6835174560547,
        267.1566467285156,
        148.74679565429688,
        321.8645324707031,
        225.14280700683594,
        320.134521484375
      ]
    ],
    "scores": [
      0.9998264908790588
    ]
  },
  "id": "045900e1-cf14-4962-923f-ffb39d264718",
  "msg": "",
  "queue": 0,
  "queue_time": 0,
  "status": 2
}

python代码调用示例：

from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

retina_face_detection = pipeline(Tasks.face_detection, 'damo/cv_resnet50_face-detection_retinaface')
img_path = 'https://guoxiuzhi.com/test/images/retina_face_detection.jpg'
result = retina_face_detection(img_path)
print(f'face detection output: {result}.')