[论文解读] NPCFace: A Negative-Positive Cooperation Supervision for Training Large-scale Face Recognition.

Dan Zeng, Hailin Shi+4 more|arXiv (Cornell University)|Jul 20, 2020

Face recognition and analysis参考文献 31被引用 9

一句话总结

NPCFace 提出了一种新颖的损失函数，显式建模了人脸识别训练中正样本边距与负样本边距之间的协作关系，从而提升了困难样本的性能，尤其是在低误接受率（FAR）场景下。通过采用样本级协作机制，联合优化类内（正样本）和类间（负样本）困难样本的边距，该方法在大规模基准测试中实现了最先进（SOTA）的准确率。

ABSTRACT

Deep face recognition has made remarkable advances in the last few years, while the training scheme still remains challenging in the large-scale data situation where many hard cases occur. Especially in the range of low false accept rate (FAR), there are various hard cases in both positives ($ extit{i.e.}$ intra-class) and negatives ($ extit{i.e.}$ inter-class). In this paper, we study how to make better use of these hard samples for improving the training. The existing training methods deal with the challenge by adding margins in either the positive logit (such as SphereFace, CosFace, ArcFace) or the negative logit (such as MV-softmax, ArcNegFace, CurricularFace). However, the correlation between hard positive and hard negative is overlooked, as well as the relation between the margin in positive logit and the margin in negative logit. We find such correlation is significant, especially in the large-scale dataset, and one can take advantage from it to boost the training via relating the positive and negative margins for each training sample. To this end, we propose an explicit cooperation between positive and negative margins sample-wisely. Given a batch of hard samples, a novel Negative-Positive Cooperation loss, named NPCFace, is formulated, which emphasizes the training on both the negative and positive hard cases via a cooperative-margin mechanism in the softmax logits, and also brings better interpretation of negative-positive hardness correlation. Besides, the negative emphasis is implemented with an improved formulation to achieve stable convergence and flexible parameter setting.We validate the effectiveness of our approach on various benchmarks of large-scale face recognition and outperform the previous methods especially in the low FAR range.

研究动机与目标

• 为解决在包含大量困难正样本和负样本的大规模数据集上训练深度人脸识别模型的挑战。
• 克服现有方法将正样本边距与负样本边距独立处理的局限，忽略了二者之间的相关性。
• 提升模型在低误接受率（FAR）范围内的泛化能力和鲁棒性，该范围正是困难样本最为关键的区域。
• 开发一种显式建模困难正样本与困难负样本之间相互作用的训练方案，以实现更优的优化效果。

提出的方法

• 提出一种基于样本的负-正协作（NPCFace）损失，联合优化类内（正样本）和类间（负样本）困难样本的边距。
• 在Softmax logits中引入协作边距机制，其中正样本的边距会根据对应负样本的难度动态调整。
• 采用改进的负样本强调公式，以稳定训练过程并支持灵活的超参数调节。
• 采用基于批量的训练策略，识别并优先对困难样本进行协作边距学习。
• 应用一种边距自适应规则，将正负样本对的难度相关联，从而增强特征的判别能力。
• 将NPCFace损失集成到标准人脸识别框架中，无需修改网络结构，支持即插即用的部署方式。

实验结果

研究问题

• RQ1 在大规模设置下，困难正样本与困难负样本之间的相关性如何影响人脸识别性能？
• RQ2 联合优化正样本与负样本边距是否能提升泛化能力，尤其是在低FAR范围内？
• RQ3 显式建模正负边距协作关系对训练稳定性与收敛性有何影响？
• RQ4 所提出的协作边距机制与现有方法中独立处理边距的策略相比，表现如何？

主要发现

• NPCFace在大规模人脸识别基准测试中实现了最先进性能，尤其在低误接受率（FAR）范围内表现突出。
• 通过协作边距学习利用正负样本之间的相关性，显著提升了困难正样本与负样本的识别准确率。
• 改进的负样本强调公式实现了更稳定的训练过程，并支持灵活的超参数控制，减少了训练不稳定性。
• 协作边距机制带来了更优的特征判别能力，表现为类间分离更紧密、类内紧凑性更强。
• 在多个大规模数据集上，该方法在低FAR评估设置下优于先前的最先进方法，如ArcFace、CurricularFace和ArcNegFace。