Implementation for "Continual Nuclei Segmentation via Prototype-wise Relation Distillation and Contrastive Learning".
utils/loss.py - PrototypeWiseRelationDistillationLoss: Prototype-wise Relation Distillationutils/loss.py - PrototypeWiseContrastiveLoss: Prototype-wise Contrastive Learning
python run.py \
--logdir $logdir \ # path to save log and checkpoint
--dataset monusac \ # dataset name
--name $name \ # experiment name
--task 1-1 \ # incremental task type
--step 0 \ # incremental step
--step_ckpt $step_ckpt \ # path to load checkpoint, invalid for step==0
--gpu $gpu # CUDA_VISIBLE_DEVICES=$gpu@article{wu2023continual,
title={Continual Nuclei Segmentation via Prototype-wise Relation Distillation and Contrastive Learning},
author={Wu, Huisi and Wang, Zhaoze and Zhao, Zebin and Chen, Cheng and Qin, Jing},
journal={IEEE transactions on medical imaging},
year={2023},
publisher={IEEE}
}This repo is based on SDR