mobarakol / P-CD

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Datasets:

The repository contains the code of the paper [Preprint] Paced-curriculum distillation with prediction and label uncertainty for image segmentation published in the journal IJCARS.
Publicly available datasets can be found as below:

  1. BUS: https://scholar.cu.edu.eg/Dataset_BUSI.zip
  2. Robotic Surgery: https://endovissub2018-roboticscenesegmentation.grand-challenge.org/Data/

Commands

Testing (Common for all models)

IID (Vanilla)

python train_model.py \
    --data bus \
    --num-classes 2 \
    --batch-size 24 \
    --epochs 100 \
    --model-path outputs/bus/iid/unet.pth \
    --tb \
    --model-name UNet \
    --train-opt IID

KD

python train_model.py \
    --train-opt KD \
    --teacher-path outputs/bus/iid/unet.pth \
    --model-path "outputs/bus/kd/kd_unet.pth" \
    --model-name UNet \
    --epochs 150 \
    --data bus \
    --batch-size 24 \
    --kld-loss-temp 4.5 \
    --tb

P-CD

python train_model.py \
    --data bus \
    --batch-size 24 \
    --epochs 150 \
    --model-path "outputs/bus/pcd_unet.pth" \
    --teacher-path outputs/bus/iid/unet.pth \
    --tb \
    --model-name UNet \
    --train-opt CD \
    --cd-mode tpl \
    --beta 0.7 \
    --gamma 0.8 \
    --kld-loss-temp 4.5 \
    --opt-t 2.51

The paper can be cited as:

@article{islam2023paced,
  title={Paced-curriculum distillation with prediction and label uncertainty for image segmentation},
  author={Islam, Mobarakol and Seenivasan, Lalithkumar and Sharan, SP and Viekash, VK and Gupta, Bhavesh and Glocker, Ben and Ren, Hongliang},
  journal={International Journal of Computer Assisted Radiology and Surgery},
  pages={1--9},
  year={2023},
  publisher={Springer}
}

About


Languages

Language:Jupyter Notebook 81.5%Language:Python 18.5%