2018
DOI: 10.48550/arxiv.1801.09414
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

CosFace: Large Margin Cosine Loss for Deep Face Recognition

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
57
0

Year Published

2018
2018
2024
2024

Publication Types

Select...
2
2
1
1

Relationship

0
6

Authors

Journals

citations
Cited by 66 publications
(70 citation statements)
references
References 21 publications
0
57
0
Order By: Relevance
“…The SphereFace [4] is the first method which proposes a multiplicative angular margin penalty to strengthen extra intra-class compactness and inter-class discrepancy simultaneously. Cosface [5] directly adds cosine margin penalty to the target logit and is much easier to train than SphereFace. Arcface [6] proposes an additive angular margin and releases a cleaned version of the MS-celeb-1m which is quite beneficial for face recognition community.…”
Section: Related Workmentioning
confidence: 99%
See 4 more Smart Citations
“…The SphereFace [4] is the first method which proposes a multiplicative angular margin penalty to strengthen extra intra-class compactness and inter-class discrepancy simultaneously. Cosface [5] directly adds cosine margin penalty to the target logit and is much easier to train than SphereFace. Arcface [6] proposes an additive angular margin and releases a cleaned version of the MS-celeb-1m which is quite beneficial for face recognition community.…”
Section: Related Workmentioning
confidence: 99%
“…. Specifically, SphereFace [4] fixes m 2 = m 3 = 0, Cosface [5] fixes m 1 = 1, m 2 = 0 and Arcface [6] fixes m 1 = 1, m 3 = 0.…”
Section: Multiface Large Margin Loss Functionsmentioning
confidence: 99%
See 3 more Smart Citations