Skip to content

Latest commit

 

History

History
15 lines (14 loc) · 749 Bytes

README.md

File metadata and controls

15 lines (14 loc) · 749 Bytes

Machine Learning project branch structure

visualization part:

  • master branch: add tsne and acc-loss curves

model part:

  • resnet branch: resnet model in pytorch without pretrain
  • my_model_1 branch: first model version designed by ksc999
  • my_model_2 branch: second model version designed by ksc999

loss function part:

  • try_pytorch_loss branch: MSELoss in pytorch
  • triplet_loss branch: TripletLoss designed by ksc999
  • center_loss branch: CenterLoss designed by ksc999
  • cosface_loss brach: CosfaceLoss from other contributors in github

add and eliminate noise part:

  • noise branch: add noise, both from source and during training; use dynamic threhold method designed by ksc999 to alleviate bad influence by noise from dataset