SeetaFace人脸识别引擎包括了搭建一套全自动人脸识别系统所需的三个核心模块，即：人脸检测模块SeetaFace Detection、面部特征点定位模块SeetaFace Alignment以及人脸特征提取与比对模块SeetaFace Identification。
主要功能：
人脸检测模块(SeetaFace Detection): 采用了一种结合传统人造特征与多层感知机（MLP）的级联结构，在FDDB上达到了84.4 的召回率（100个误检时），并可在单个i7 CPU上实时处理VGA分辨率的图像。
面部特征点定位模块(SeetaFace Alignment): 通过级联多个深度模型（栈式自编码网络）来回归5个关键特征点（两眼中心、鼻尖和两个嘴角）的位置，在AFLW数据库上达到state-of-the-art的精度，定位速度在单个i7 CPU上超过200fps。
人脸识别模块(SeetaFace Identification): 采用一个9层的卷积神经网络（CNN）来提取人脸特征，在LFW数据库上达到97.1 的精度（注：采用SeetaFace人脸检测和SeetaFace面部特征点定位作为前端进行全自动识别的情况下），特征提取速度为每图120ms（在单个i7 CPU上）。
The SeetaFace Face Recognition Engine includes the three core modules required to build a fully automated face recognition system, namely the Face Detection Module SeetaFace Detection, the Face Feature Point Segment Module SeetaFace Alignment, and the Face Feature Extraction and Matching Module SeetaFace Identification.
The main function:
Face Detection Module (SeetaFace Detection): A combination of traditional artificial features and multi-layer sensor (MLP) cascade structure, in the FDDB reached 84.4 recall rate (100 false detection), and Can process VGA resolution images in real time on a single i7 CPU.
Face feature positioning module (SeetaFace Alignment): by cascading multiple depth model (stack self-coding network) to return to five key feature points (two centers, nose and two mouth) position, in the AFLW To achieve the state-of-the-art accuracy, positioning speed in a single i7 CPU more than 200fps.
Face Recognition Module (SeetaFace Identificat (2017-04-07, Java, 2KB, 下载19次)