SphereFace[2017-CVPR]

Paper-info

• title : SphereFace: Deep Hypersphere Embedding for Face Recognition[CVPR-2017]
• author : Weiyang Liu, Yandong Wen, Zhiding Yu, Ming Li, Bhiksha Raj, Le Song.
• insight : angular softmax (A-Softmax) loss.
• github : metricface-pytorch

Motivation

• prior : 人脸位于流行结构上.
• 基于角度的softmax损失函数本质是：在超球面上加入了区分约束，这与先验知识非常匹配！

 

Details

                                          

• 归一化W，同时让bias置零

                                                           

1. 优化内积 --> 优化角度
2. decision boundary更为简洁明了，人脸分布于空间超球面上！

• cos(mθ1)>cos(θ2)

                                     

     1. 为了保证cos(m*theta)单调递减，需要对theta的取值范围加以约束，为了消除这种约束，采用如下的方式来替代cos：

                                       ， 

# 根据m*theta的取值范围，找到对应的k，使得cos(m*theta)在对应的区间单调
theta = cos_theta.data.acos()
k = (self.m * theta / 3.14159265).floor()   # k.shape == theta.shape

Sphereface-pytorch

class Sphereface(nn.Module):
  
    def __init__(self, in_features, out_features, m = 4, use_gpu = False):

        super(Sphereface, self).__init__()

        self.in_features  = in_features
        self.out_features = out_features
        self.m     = m
        self.base  = 1000.0
        self.gamma = 0.12
        self.power = 1
        self.LambdaMin = 5.0
        self.iter   = 0
        self.weight = torch.FloatTensor(out_features, in_features)
        if use_gpu:
            self.weight = self.weight.cuda()
        self.weight = Parameter(self.weight)
        nn.init.xavier_uniform_(self.weight)
        
        self.mlambda = [
            lambda x: x ** 0,
            lambda x: x ** 1,
            lambda x: 2 * x ** 2 - 1,
            lambda x: 4 * x ** 3 - 3 * x,
            lambda x: 8 * x ** 4 - 8 * x ** 2 + 1,
            lambda x: 16 * x ** 5 - 20 * x ** 3 + 5 * x
        ]

    def forward(self, input, target):

        self.iter += 1
        self.lamb = max(self.LambdaMin, self.base * (1 + self.gamma * self.iter) ** (-1 * self.power))

        cos_theta = F.linear(F.normalize(input), F.normalize(self.weight))
        cos_theta = cos_theta.clamp(-1, 1)
        cos_m_theta = self.mlambda[self.m](cos_theta)
        theta = cos_theta.data.acos()
        k = (self.m * theta / 3.14159265).floor()
        phi_theta = ((-1.0) ** k) * cos_m_theta - 2 * k
        NormOfFeature = torch.norm(input, 2, 1)

        one_hot = torch.zeros(cos_theta.size(), device=input.device)
        one_hot.scatter_(1, target.view(-1, 1), 1)

        output = (one_hot * (phi_theta - cos_theta) / (1 + self.lamb)) + cos_theta
        output *= NormOfFeature.view(-1, 1)

        return output

Experiments

• Sphere-3D-CASIA_WebFace         

              

• LFW/YTF