Prepare materials
1. Service Provisioning ( new first consumer purchase, within one month of not more than 2QPS, free )
Link Address: https://common-buy.aliyun.com/?commodityCode=face_pre#/open
2. pay can be found on the console (stage Management Console is not perfect, but does not affect the test caller).
Link Address: https://face.console.aliyun.com/?spm=a2c4e.11153940.0.0.549650b0IhjgTo#/HomeIndex
Face Recognition Face Recognition
Face Recognition Based on face detection in the image or video analysis and matching techniques, to provide positioning face detection, face recognition and face attributes other than an independent service module. Can provide high-performance online service API for developers and businesses, applied to the face AR, face recognition and certification, large-scale search of the human face, photo management and other scenes.
Product features include:
Face detection and location
It was identified whether a face image and the face region are given, for face to face return detection coordinates detected rectangle. Can support up to thousands of simultaneous detection of human faces, a plane support face detection 360 degree rotation, supports a maximum of 90 degrees left and right side face detection. Meanwhile millisecond extracted face image features key, identifying key face 105 is positioned.
Face Recognition Properties
In the face detection based on the recognition of each person's face to detect gender (male / female), age, facial expression (smiling / laughing), glasses (wearing / not wearing) four attributes. Return to face the 1024 feature-dimensional depth study, based on the characteristics and features in accordance with the rules of comparison, users can achieve high performance of face recognition. Our algorithm in the open test set lfw reached 99.58% recognition accuracy.
Face alignment
Two pictures based on user input, can detect human faces in the two pictures, and pick two pictures of the largest face, and determines whether the same person. Returning the two men face rectangle coordinates, than the confidence level, confidence threshold and the error rate is different.
Face Finding 1: N
基于输入的图像或视频，与注册库比对，实现1：N的人脸识别。适用于人脸登录、VIP人脸识别等无需刷卡验证的场景。提供sdk调用方式，支持无交互式或动作活体检测，支持多平台，CPU、GPU计算模式，灵活部署，数千万人脸样本收集，CNN网络模型的不断训练和调优，兼顾效率和精度的身份对比网络。
应用体验 - ET体验馆
https://data.aliyun.com/ai?spm=a2c0j.8207365.813079.11.69162fd59RHeiH#/face-comparison
参考资料
阿里云新版人脸识别综述
https://yq.aliyun.com/articles/711726?spm=a2c4e.11153940.0.0.722349a0CuB3Yf
阿里云人脸识别 1:N 使用简明示例
https://yq.aliyun.com/articles/704556?spm=a2c4e.11153940.0.0.6e673d1eyRvTWv
阿里云新版人脸识别 C# 使用示例教程
https://yq.aliyun.com/articles/710449?spm=a2c4e.11153940.0.0.6e673d1eyRvTWv