Skip to content

「Pytorch」Try to resolve Face Alignment problems by LinearModel,ResNet18,MobileNetV2(PFLD)

Notifications You must be signed in to change notification settings

ideask/Face-Alignment

Repository files navigation

Face Alignment

Introduction

Using Pytorch as a framework, based on Linear model,ResNet18 or MobileNetV2

1. Train Linear Model:

  • Data preparation:

    • Run python ./Data/ODATA/linear.py
  • Training steps:

    • Run tensorboard --logdir=/home/kenny/Desktop/Face-Alignment/CheckPoints/tensorboard_linear &
    • Run python Train_linear.py -h get usage
    • Run default parms python Train_linear.py
    • Checkpoint checkpoint_epoch_x.pth.tarin./CheckPoints/snapshot_linear/
    • You can get training log file from ./CheckPoints/train_linear.logs
  • Testing steps:

    • Run python Test_linear.py -h get usage
    • Run default parms python Test_linear.py

2. Train ResNet18 Model:

  • Data preparation:

    • Run python ./Data/ODATA/resnet.py
  • Training steps:

    • Run tensorboard --logdir=/home/kenny/Desktop/Face-Alignment/CheckPoints/tensorboard_resnet &
    • Run python Train_resnet.py -h get usage
    • Run default parms python Train_resnet.py
    • Checkpoint checkpoint_epoch_x.pth.tarin./CheckPoints/snapshot_resnet/
    • You can get training log file from ./CheckPoints/train_resnet.logs
  • Testing steps:

    • Run python Test_resnet.py -h get usage
    • Run default parms python Test_resnet.py

3. Train MobileNetV2 Model(refer to PFLD):

  • Data preparation:

    • Run python ./Data/ODATA/pfld.py
  • Training steps:

    • Run tensorboard --logdir=/home/kenny/Desktop/Face-Alignment/CheckPoints/tensorboard_pfld &
    • Run python Train_pfld.py -h get usage
    • Run default parms python Train_pfld.py
    • Checkpoint checkpoint_epoch_x.pth.tarin./CheckPoints/snapshot_pfld/
    • You can get training log file from ./CheckPoints/train_pfld.logs
  • Testing steps:

    • Run python Test_pfld.py -h get usage
    • Run default parms python Test_pfld.py

Result

Predict landmarks:Green Points
Ground Truth landmarks:Red Points

Linear Model:

  • Loss

  • Predict

ResNet18:

  • Loss

  • Predict

MobileNetV2(refer to PFLD):

  • Loss

  • Predict

Reference

About

「Pytorch」Try to resolve Face Alignment problems by LinearModel,ResNet18,MobileNetV2(PFLD)

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages