Skip to content

Jingnan-Jia/lobeseg

Repository files navigation

Multi-task Semi-supervised Learning for Lobe Segmentation

Introduction for several directories with their specific functions:

  • data (training datasets are saved here)
  • futils (common used functions and models including building models, compute metrics)
  • logs (save monitor metrics during training)
  • models (save trained models)
  • results (save the training/validation/testing results including Dice, MSD, Hausdorff distance, false positive, etfc.)

Introduction for each files

  • Use python train_ori_fit_rec_epoch.py to train model.
  • Use write_preds_save_dice.py to evaluate the trained model.
  • Modify set_parameters.py to set custom parameters.
  • Scipts files script* are used to submit job to HPC cluster.
  • plot_curve* are used to plot training loss curve.

About

new repositary for multi-task learning

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published