cavaface: A Pytorch Training Framework for Deep Face Recognition

.

By Yaobin Li and Liying Chi

Introduction

This repo provides a high-performance distribute parallel training framework for face recognition with pytorch, including various backbones (e.g., ResNet, IR, IR-SE, ResNeXt, AttentionNet-IR-SE, ResNeSt, HRNet, etc.), various losses (e.g., Softmax, Focal, SphereFace, CosFace, AmSoftmax, ArcFace, ArcNegFace, CurricularFace, Li-Arcface, QAMFace, etc.), various data augmentation(e.g., RandomErasing, Mixup, RandAugment, Cutout, CutMix, etc.) and bags of tricks for improving performance (e.g., FP16 training(apex), Label smooth, LR warmup, etc)

Features

<details open> <summary>(click to collapse)</summary>

• Backbone
  • [x] ResNet(IR-SE)
  • [x] ResNeXt
  • [x] DenseNet
  • [x] MobileFaceNet
  • [x] MobileNetV3
  • [x] EfficientNet
  • [x] ProxylessNas
  • [x] GhostNet
  • [x] AttentionNet-IRSE
  • [x] ResNeSt
  • [x] ReXNet
  • [x] MobileNetV2
  • [x] MobileNeXt
• Attention Module
  • [x] SE
  • [x] CBAM
  • [x] ECA
  • [x] GCT
• Loss
  • [x] Softmax
  • [x] SphereFace
  • [x] AMSoftmax
  • [x] CosFace
  • [x] ArcFace
  • [x] Combined Loss
  • [x] AdaCos
  • [x] SV-X-Softmax
  • [x] CurricularFace
  • [x] ArcNegFace
  • [x] Li-Arcface
  • [x] QAMFace
  • [x] Circle Loss
• Parallel Training
  • [x] DDP
  • [x] Model Parallel
• Automatic Mixed Precision
  • [x] AMP
• Optimizer
  • [x] LRScheduler(faireq,rwightman)
  • [x] Optim(SGD,Adam,RAdam,LookAhead,Ranger,AdamP,SGDP)
  • [x] ZeRO
• [Data Augmentation
  • [x] RandomErasing
  • [x] Mixup
  • [x] RandAugment
  • [x] Cutout
  • [x] CutMix
  • [x] Colorjitter
• Distillation
  • [ ] KnowledgeDistillation
  • [ ] Multi Feature KD
• Bag of Tricks
  • [x] Label smooth
  • [x] LR warmup

</details>

Installation

See INSTALL.md.

Quick start

See GETTING_STARTED.md.

Model Zoo and Benchmark

See MODEL_ZOO.md.

License

cavaface is released under the MIT license.

Acknowledgement

• This repo is modified and adapted on these great repositories face.evoLVe.PyTorch, CurricularFace, insightface and imgclsmob
• The evaluation tools is developed by Charrin

Contact

cavallyb@gmail.com