论文阅读（9）：AdapticeFace: Adaptive Margin and Sampling for Face Recognition
本文是对Margin based Softmax Loss的改进。主要针对人脸数据集类别不平衡的问题，以及难类别/难样本的挖掘，让训练过程变得高效。 一、针对的问题 SOTA Margin based Softmax Loss都有一个前提假设：即所有的类别都有足够多的样本来描述该类别的分布，因此一个预设的常量margin就足以缩小每一个类别的类内差异。实际上人脸数据集的样本很不平衡，类别的
相关文章
- 1. 论文阅读（9）：AdapticeFace: Adaptive Margin and Sampling for Face Recognition
- 2. AdaptiveFace: Adaptive Margin and Sampling for Face Recognition 阅读笔记
- 3. 论文阅读笔记：ArcFace: Additive Angular Margin Loss for Deep Face Recognition
- 4. FaceNet: A Unified Embedding for Face Recognition and Clustering（论文阅读笔记）
- 5. 论文阅读 FaceNet: A Unified Embedding for Face Recognition and Clustering
- 6. 文献阅读 - Deep Face Recognition
- 7. 文献阅读 - FaceNet: A Unified Embedding for Face Recognition and Clustering
- 8. SphereFace: Deep Hypersphere Embedding for Face Recognition论文阅读
- 9. 【ArcFace/Insightface论文中文版】ArcFace: Additive Angular Margin Loss for Deep Face Recognition
- 10. 论文阅读（5）：Age invariant face recognition and retrieval by coupled auto-encoder networks
- 更多相关文章...
- • RSS 阅读器 - RSS 教程
- • C# 文本文件的读写 - C#教程
- • RxJava操作符（七）Conditional and Boolean
- • JDK13 GA发布：5大特性解读
相关标签/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
欢迎关注本站公众号,获取更多信息
相关文章
- 1. 论文阅读（9）：AdapticeFace: Adaptive Margin and Sampling for Face Recognition
- 2. AdaptiveFace: Adaptive Margin and Sampling for Face Recognition 阅读笔记
- 3. 论文阅读笔记：ArcFace: Additive Angular Margin Loss for Deep Face Recognition
- 4. FaceNet: A Unified Embedding for Face Recognition and Clustering（论文阅读笔记）
- 5. 论文阅读 FaceNet: A Unified Embedding for Face Recognition and Clustering
- 6. 文献阅读 - Deep Face Recognition
- 7. 文献阅读 - FaceNet: A Unified Embedding for Face Recognition and Clustering
- 8. SphereFace: Deep Hypersphere Embedding for Face Recognition论文阅读
- 9. 【ArcFace/Insightface论文中文版】ArcFace: Additive Angular Margin Loss for Deep Face Recognition
- 10. 论文阅读（5）：Age invariant face recognition and retrieval by coupled auto-encoder networks