• DOI:10.1109/CVPR.2018.00552
Corpus ID: 68589

CosFace: Large Margin Cosine Loss for Deep Face Recognition

@article{Wang2018CosFaceLM,
  title={CosFace: Large Margin Cosine Loss for Deep Face Recognition},
  author={H. Wang and Yitong Wang and Z. Zhou and Xing Ji and Zhifeng Li and Dihong Gong and Jingchao Zhou and Wenyu Liu},
  journal={2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2018},
  pages={5265-5274}
}

• H. Wang, Yitong Wang, +5 authors Wenyu Liu
• Published 2018
• Computer Science
• 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition

Face recognition has made extraordinary progress owing to the advancement of deep convolutional neural networks (CNNs. [...] Key Method More specifically, we reformulate the softmax loss as a cosine loss by L2 normalizing both features and weight vectors to remove radial variations, based on which a cosine margin term is introduced to further maximize the decision margin in the angular space. As a result, minimum intra-class variance and maximum inter-class variance are achieved by virtue of normalization and…Expand

View on IEEE

arxiv.org

Save to Library

Create Alert

Cite

Launch Research Feed