ArcFace: Additive Angular Margin Loss for Deep Face Recognition
Citations
30 citations
Cites methods from "ArcFace: Additive Angular Margin Lo..."
...The bounding box of the face is extracted using the RetinaFace (Deng et al., 2019b)....
[...]
...Identities are extracted using ArcFace (Deng et al., 2019a)....
[...]
30 citations
30 citations
Cites background or methods from "ArcFace: Additive Angular Margin Lo..."
...In the area of face recognition, ArcFace [11] reached a precision of 99....
[...]
...It should be mentioned that, ArcFace [11] achieved a precision of 99....
[...]
...The faces are detected by the SSH model [39], and then recognized by the ArcFace model [11]....
[...]
...We train a head classifier based on the ArcFace model [11]....
[...]
...The state-of-art method, ArcFace [11] achieved a face verification accuracy of 99....
[...]
30 citations
Cites background or methods from "ArcFace: Additive Angular Margin Lo..."
...As pre-processing, we normalize a face image to 112× 112 by warping a face-region using five facial points from two eyes, nose and two corners of mouth [6,22,40]....
[...]
...The recent adoption [6,7,22,32,37,39,40,41] of Convolutional Neural Networks (CNNs) has dramatically increased recognition accuracy....
[...]
...A backbone network is ResNet-100 [11] that is used in the recent works [6,15]....
[...]
...1 ArcFace [6] 99....
[...]
...The computed embedding vectors and the weight vectors of the linear classifier are L2-normalized and trained by the ArcFace [6]....
[...]
30 citations
Cites background or methods from "ArcFace: Additive Angular Margin Lo..."
...In face recognition, it is very important to perform open-set evaluation [28, 42, 7], i....
[...]
...To achieve better performance, large CNNs like ResNet [7] or AttentionNet [46] are usually employed, which makes them hard to deploy on mobile and embedded devices....
[...]
...There are many kinds of network architectures [28, 3, 41] and several loss functions [7, 46] for face recognition....
[...]
...Specifically, rather than the traditional softmax loss, face recognition is usually supervised by margin-based softmax losses [28, 24, 42, 47, 7, 46], metric learning losses [37] or both [39]....
[...]
...Without loss of generality, we use SEResNet50-IR [7] as the teacher model, which was trained by SV-AMSoftmax loss [46]....
[...]
References
123,388 citations
33,597 citations
30,843 citations
13,268 citations
10,447 citations