cvqluu/Angular-Penalty-Softmax-Losses-Pytorch 开源项目中文解读：技术栈、适用场景与上手建议

499

MIT License

项目介绍

cvqluu/Angular-Penalty-Softmax-Losses-Pytorch 是一个托管在 GitHub 上的开源项目，主要使用 Python 编写，采用 MIT License 许可证，目前获得 499 个 Star、91 次 Fork。

仓库全名: cvqluu/Angular-Penalty-Softmax-Losses-Pytorch
开源许可证: MIT License
Star 数: 499
Fork 数: 91
未关闭 Issue: 15

README 摘要

Angular Penalty Softmax Losses Pytorch Concise Pytorch implementation of the Angular Penalty Softmax Losses presented in: ArcFace: https://arxiv.org/abs/1801.07698 [1] SphereFace: https://arxiv.org/abs/1704.08063 [2] CosFace/Additive Margin: https://arxiv.org/abs/1801.09414 [3] / https://arxiv.org/abs/1801.05599 [4] (Note: the SphereFace implementation is not exactly as described in their paper but instead uses the…

完整内容见 GitHub 原始 README 或本页。

常见问题

Angular-Penalty-Softmax-Losses-Pytorch 用什么语言开发？

主要使用 Python 编写，采用 MIT License 许可证，目前有 499 个 Star。

Angular-Penalty-Softmax-Losses-Pytorch 可以免费商用吗？

如何获取 Angular-Penalty-Softmax-Losses-Pytorch 的源码？

可以直接克隆仓库：git clone https://github.com/cvqluu/Angular-Penalty-Softmax-Losses-Pytorch.git；安装与运行步骤见仓库 README。

Angular-Penalty-Softmax-Losses-Pytorch 最近还在维护吗？

仓库最近一次更新为 2023年12月13日，当前有 15 个未关闭 Issue；Star 数为 499，可作为社区活跃度的参考。

关于

Angular penalty loss functions in Pytorch (ArcFace, SphereFace, Additive Margin, CosFace)

描述取自 GitHub 官方 API，中文内容由 GitRead 中文社区编辑团队整理。

简要分析

热度分

1180

价值分

375

活跃状态

活跃

am-softmax arcface embedding face-recognition face-verification fashion-mnist fmnist-dataset loss-function loss-functions metric-learning normface pytorch speaker-recognition sphereface

GitHub

语言占比

Python

贡献者

cvqluu

README

Angular Penalty Softmax Losses Pytorch

Concise Pytorch implementation of the Angular Penalty Softmax Losses presented in:

ArcFace: https://arxiv.org/abs/1801.07698 [1]
SphereFace: https://arxiv.org/abs/1704.08063 [2]
CosFace/Additive Margin: https://arxiv.org/abs/1801.09414 [3] / https://arxiv.org/abs/1801.05599 [4]

(Note: the SphereFace implementation is not exactly as described in their paper but instead uses the ‘trick’ presented in the ArcFace paper to use arccosine instead of the double angle formula)


from loss_functions import AngularPenaltySMLoss

in_features = 512
out_features = 10 # Number of classes

criterion = AngularPenaltySMLoss(in_features, out_features, loss_type='arcface') # loss_type in ['arcface', 'sphereface', 'cosface']

# Forward method works similarly to nn.CrossEntropyLoss
# x of shape (batch_size, in_features), labels of shape (batch_size,)
# labels should indicate class of each sample, and should be an int, l satisying 0 <= l < out_dim
loss = criterion(x, labels) 
loss.backward()

Experiments/Demo

There are a simple set of experiments on Fashion-MNIST [2] included in train_fMNIST.py which compares the use of ordinary Softmax and Additive Margin Softmax loss functions by projecting embedding features onto a 3D sphere.

The experiments can be run like so:

python train_fMNIST.py --num-epochs 40 --seed 1234 --use-cuda

Which produces the following results:

Baseline (softmax)

softmax

Additive Margin Softmax/CosFace

cosface

ArcFace

arcface

TODO: fix sphereface results

[1] Deng, J. et al. (2018) ‘ArcFace: Additive Angular Margin Loss for Deep Face Recognition’. Available at: http://arxiv.org/abs/1801.07698.

[2] Liu, W. et al. (2017) ‘SphereFace: Deep hypersphere embedding for face recognition’, in Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017, pp. 6738–6746. doi: 10.1109/CVPR.2017.713.

[3] Wang, H. et al. (2018) ‘CosFace: Large Margin Cosine Loss for Deep Face Recognition’. Available at: http://openaccess.thecvf.com/content_cvpr_2018/papers/Wang_CosFace_Large_Margin_CVPR_2018_paper.pdf (Accessed: 12 August 2019).

[4] “Additive Margin Softmax for Face Verification.” Wang, Feng, Jian Cheng, Weiyang Liu and Haijun Liu. IEEE Signal Processing Letters 25 (2018): 926-930.

[5] “Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms.” Han Xiao, Kashif Rasul, Roland Vollgraf. arXiv:1708.07747

Angular-Penalty-Softmax-Losses-Pytorch