[h1]人脸识别API/SDK 汇总 [/h1][h1]1、face.com[/h1]以色列公司，某年六月时被Facebook收购，同时暂停了API服务，之前测试过他们的服务，基本上是了解到的应用中做得最牛的了。 [h1] 2、orbe[/h1]Orbeus由麻省理工学院和波士顿大学的几个科学家联合创立，他们致力于让Orbeus实现能从照片或视频中识别出所有内容。 测试：结果同样非常另人满意！ [h1] 3、Face++[/h1]Face++.com 是一个提供免费人脸检测、人脸识别、人脸属性分析等服务的云端服务平台。Face++是北京旷视科技有限公司旗下的全新人脸技术云平台，在黑马大赛中，Face++获得年度总冠军，已获得联想之星投资。 测试：在国内公开的此类服务中属佼佼者！ [h1]4、skybiometry[/h1] 测试：结果也相当不错！ 5、Luxand 测试：结果一般，就不图了！ 6、lambdal 7、 betaface 8、Facedetection 列出的SDK list: CompanySDKOnline DocumentationTechnology ACSYSCanada AcsysFRS SDKHNet(Holographic/Quantum Neural Technology) BetafaceGermanyBetaface Face Detection and Recognition SDKhttp://www.betaface.com/Betaface%20SDK.pdfCognitecGermanyFaceVACSSDKhttp://www.cognitec-ag.de/documentation/frsdk_publicNeural Nets (I assume)L-1USAFaceITSDKDetection: Pattern MatchingRecognition: "Local Feature Analysis"Luxand USAFaceSDKFaceSDK Documentation NEUROtechnologija Lithuania VeriLook SDKVisage Technologies ABSwedenvisage|SDKDetection: Active Appearance ModelsTracking: Template Matching + Kalman Filter [h1]9、相关文章：[/h1][h1]反人脸识别[/h1][h1]人脸识别App集合[/h1]

开源的人脸识别没有一个靠谱的。OpenCV等自带的很弱，face++等是服务接口，不是开源代码。 这两年随着深度学习的流行，人脸识别技术提高一大截，主要是归一化和特征提取部分。虽然很多机构和公司都发了准确率刷榜文章，但没有一家release出源码和模型。 以前常用的ASM、Gabor、LBP等相比落后很多了。所以如果是三年前的技术，基本也落后了。 所以结论是，想要有靠谱的技术，只能自己做。人脸检测、特征点定位归一化、人脸特征与识别模型，每一块都需要至少一个月时间，不熟悉的话，几个月也正常。

那我就讲个实话吧，甭管你说LWF识别率99%以上，面对大规模人脸识别的时候，统统都很渣，我就问你们业内的人承不承认？这是知乎，不是渣浪等放卫星般的媒体，能踏实点嘛！

引用自Quora： s@lambdal.com. Happy Hacking! -Stephen[*]Face (Detection) - A computer vision api for facial recognition and facial detection that is a perfect Face.com | Coming Soon replacement. We currently have a free api for face detection. [/*][*]Animetrics Face Recognition - The Animetrics Face Recognition API can be used to detect human faces in pictures. Information on facial features or "landmarks" is returned as coordinates on the image.. Animetrics Face Recognition will also detect and return the orientation, or "pose" of faces along 3 axes.[/*][*]Skybiometry Face Detection and Recognition - An easy to use Face Detection and Recognition API. You must have an application created in your account at SkyBiometry to use it (sign up at if you don't have account yet).[/*][*]ImageVision Face Detection - Beta Release A better Face detection service! ImageVision is a computer vision company improving technology that determines the locations and sizes of human faces in arbitrary (digital) images.[/*][*]Face++ - Face++ uses the cutting-edge technology of computer vision and data mining to provide 3 core vision services (Detection, Recognition, and Analysis). With the service and huge database of celebrity from Face++, the developers can apply the face technology into their own websites, mobile Apps, and smart TVs, increasing incredible user experience[/*][*]FaceMark - FaceMark is a powerful API for facial feature detection. It finds 68 points for a frontal face and 35 for a profile one. FaceMark detects landmarks for faces on the image specified by URL or uploaded as a file and produces JSON output containing a vector of facial landmarks and orientation for each face found.[/*][*]EmoVu by Eyeris - Artificially Intelligent emotion recognition software that allows cameras to “look back” and read people’s facial micro-expressions, gender and age group in real time as they watch videos on their computers or mobile devices.[/*][*]Face and scene recognition by http://Rekognition.com - Face.com | Coming Soon alternative! Our fast, robust and scalable rekognition engine can do facial detection, crawling, recognition, scene understanding! It can be automatically trained using images and tags on Facebook![/*][*]FaceRect - FaceRect is a powerful and free API for face detection. It finds faces (both frontal and profile) on the image specified by URL or uploaded as a file and is able to find multiple faces on a single photo, producing JSON output with a bounding box for each face found.[/*][*]Infatics Face Detection - Simple face detection API[/*][*]OpenCV Face Recognizer - OpenCV (Open Source Computer Vision Library: OpenCV | OpenCV) is an open-source BSD-licensed library that includes several hundreds of computer vision algorithms.[/*][*]Libface - Libface is a library aimed at bringing face recognition technology to the open source community. It is written in C++ and maintained on Sourceforge. The library uses OpenCV 2.0 and aims to be a middleware for developers that don’t have to include any OpenCV code in order to use face recognition and face detection detection.[/*][*]Automatic naming of characters in video - The objective of this work is to label television or movie footage with the names of the people present in each frame of the video.[/*][*]CCV - a modern computer vision library[/*][*]OpenBR - Open Source Biometric Recognition[/*][*]Flandmark - open source implementation of facial landmark detector[/*][*]Computer Expression Recognition Toolbox (CERT) - an end-to-end system for fully automated facial expression recognition that operates in real-time.[/*][*]Nviso 3D facial imaging technology - Analyze human emotions from facial expressions. More direct and automatic than by any other method.[/*][*]FaceReader - FaceReader is the world's first tool that is capable of automatically analyzing facial expressions, providing users with an objective assessment of a person’s emotion.[/*][*]Affdex - Affdex reads emotional states such as liking and attention from facial expressions using a webcam... to give marketers faster, more accurate insight into consumer response to brands and media.[/*][*]Faceware - Analyzer takes in a video of an actor’s facial performance and produces an .IMPD file for use in Retargeter™. It accomplishes this through the combination of simple user input on the front end, and complex computer vision algorithms on the back end.[/*][*]Face detection in iOS[/*][*]Face-Recognition-SDK - add face recognition to your apps[/*][*]Oddcast face detection API - The Face Tracking API allows Flash developers to harness advanced technology previously available only in high end video games.[/*][*]BioID - world's first webcam-based personal recognition[/*][*]Betaface API - face detection and recognition[/*][*]Discrete Area Filters Face Detector - detect 15 face points, 4 face parts, multiple face, occluded faces[/*][*]Face detection using Support Vector Machine - this program is the clone of 'Face Detection System' in MATLAB but instead of Neural Networks, it is based on SVM.[/*][*]fdlib - is a face detection library for C/C++ and Matlab. It allows you to detect frontal faces in images.[/*][*]Visage - a human computer interface that aims to replace the traditional mouse with the face. Using a webcam and Visage the movement of the face becomes the movement of the mouse pointer. Left/right Eye blinks fire left/right mouse click events.[/*][*]Face tracking Project (Carnegie Mellon) - the Advanced Multiple Processing (AMP) offers face tracking using color matching combined with deformable templates. Both DLL for Windows and static library for Linux [/*][*]Real-time face detection program - Demo from the Fraunhofer institute IIS, Germany. Shows face tracking and detection using edge orientation matching.[/*][*]Evaluation of Face Recognition Algorithms - Resource for all researchers developing face recognition algorithms from Colorado State University. It provides a standard set of well known algorithms and established experimental protocols. [/*][*]Computer Vision Source Code - useful collection of image processing code[/*][*]Acsys biometrics SDK - allows third-party developers to implement advanced facial biometric technology with their own security applications[/*][*]Cognitec SDK - develops market-leading face recognition technologies and applications for enterprise and government customers around the world.[/*][*]KeyLemonFaceSDK - integrates recognition technology into major operating systems.[/*][*]FaceIT SDK -[/*][*]FaceSDK - FaceSDK enables Microsoft Visual C++, C#, VB, Java, and Delphi developers to build 32-bit and 64-bit applications for Web, Windows, Linux, and MacOS X with face recognition and face-based biometric identification functionality[/*][*]Verilook SDK - allows rapid development of biometric applications using functions from the VeriLook algorithm that ensure fast and reliable face identification.[/*][*]Beyond Reality Face SDK - calculates the position and 3D angles of a face in a single picture of a webcam stream. This information can be used to place 3D objects onto the image or to control an app via head movements[/*][*]Sightcorp - cutting edge face analysis technologies, targeting individual or crowd analysis[/*][*]Visage FaceDetect SDK - contains powerful techniques to find faces and facial features in still images in form of a well-documented C++ Software Development Kit[/*][*]Microsoft Research Face SDK Beta - integrates the latest face technologies from Microsoft research teams.[/*][*]How To: Kinect for Windows SDK Face Recognition[/*][*]Bayometric FaceIt Face Recognition SDK - combines traditional facial geometry techniques with skin biometrics[/*][*]FacePhi FaceRecognition SDK - consists of a set of libraries developerd for .NET and Silverlight[/*][*]360Biometrics Face SDK - highly advanced face recognition system that is designed to work with the embedded or linked face images in your existing database.[/*][*]Hunter TrueID - web-enabled face recognition[/*][*]IMRSV - a real world, real time perceptive computing software that measures gender, age, attention time, dwell, glances and more for up to 25 people, 25 ft away using a basic webcam. It has a REST API as well.[/*][*]Bob - Bob is a signal-processing and machine learning toolbox originally developed by the Biometrics Group at Idiap, in Switzerland. Bob is primarily developed through GitHub.[/*][*]KeyLemon - KeyLemon provides a new way to access your computer. Instead of typing your password, you just need to smile to the webcam and KeyLemon will recognize you[/*][*]Flash AS3 face detection and recognition library - Oskar Wicha’s ActionScript 3 library allow us to do both things using Eigenfacesconcept, basically some patterns which should match a human face.[/*][*]Stupeflix - Return an array of positions of detected faces, with type and confidence.[/*][/list]

开源人脸识别软件，到目前为止还没有什么精度比较好的。 OpenCV里面的人脸检测及其相关的haar detector还是蛮不错的，但想搞靠谱实用还是的自己重新train。 人脸识别最好的几家公司都在07年的frvt（google收购的NV参与了这届，综合性能第一）和后来的10年的mbe（google收购的pittpatt参加了这届，性能一般，但他们的人脸检测、姿态估计似乎很好，cmu的人马，专利不少）网站上了。看看就知道最好的人脸识别技术能达到什么程度了

最吊的还是云从科技，真的，拿实力说话

openface TensorFlow 的实现：davidsandberg/facenet 训练好的的模型参数 ，LFW 上的正确率 99.3%

楼主可以关注下这个人脸检测的benchmark：Fine-grained Evaluation on Face Detection in the Wild 里面整理了一些state-of-the-art的人脸检测方法： 应该够用了吧。

看了下很多人说的是face detection而非face recognition：前者是检测，后者才是识别。目前face++的确是人脸方面做得很不错的在线API，其中的“face search”和百度识图，google以图搜图等基本相似，但是精度没那么高，最终都可以看做是face recognition的一种变体face classification。此外基于opencv开发的各种人脸识别算法也很多，但是成型的开源软件太少，我想一方面是因为人脸识别这个课题依赖于应用场景或者说是测试集，单纯一个识别软件不可能通吃；另一方面是很多机构或者公司基于一套算法开发一个识别软件基本都用于商业化了，不会开源。最后想说如今人脸识别的算法实在太多太多了，拿一个算法基于opencv或者matlab做出一个软件自己玩玩还是挺ok的，至于开源与否，我觉得关系不大。

国际人脸识别评估集LFW榜， face++ 99.5% , 商汤 99.53% ， 腾讯 99.65% , 百度 99.77%，中科奥森 99.77% , 中科云从 99.5%， 颜鉴99.64%，北京飞搜科技 99.4% ,宇泛智能 99%，