This is an official implementation of the paper "Replay Without Saving: Prototype Derivation and Distribution Rebalance for Class-Incremental Semantic Segmentation", accepted by IEEE TPAMI. 📝 Paper

This repository has been tested with the following environment:

• CUDA (11.3)
• Python (3.8.13)
• Pytorch (1.12.1)
• Pandas (2.0.3)

conda create -n star python=3.8.13
conda activate star
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch
conda install pandas==2.0.3

We use augmented 10,582 training samples and 1,449 validation samples for PASCAL VOC 2012. You can download the original dataset in here. To train our model with augmented samples, please download labels of augmented samples ('SegmentationClassAug') and file names ('train_aug.txt'). The structure of data path should be organized as follows:

└── ./datasets/PascalVOC2012
    ├── Annotations
    ├── ImageSets
    │   └── Segmentation
    │       ├── train_aug.txt
    │       └── val.txt
    ├── JPEGImages
    ├── SegmentationClass
    ├── SegmentationClassAug
    └── SegmentationObject
    
    

We use 20,210 training samples and 2,000 validation samples for ADE20K. You can download the dataset in here. The structure of data path should be organized as follows:

└── ./datasets/ADE20K
    ├── annotations
    ├── images
    ├── objectInfo150.txt
    └── sceneCategories.txt

We use 2975 training samples and 500 validation samples for ADE20K. You can download the dataset in here ("gtFine_trainvaltest.zip (241MB) [md5]" and "leftImg8bit_trainvaltest.zip (11GB) [md5]"). The structure of data path should be organized as follows:

└── ./datasets/ADE20K
    ├── leftImg8bit
    ├── gtFine
    ├── fine_train.txt
    └── fine_val.txt

To train STAR-Lite and STAR-Basic models, you can directly run the .sh files located in the ./scripts/ directory.

To evaluate on the PASCAL VOC 2012 dataset, execute the following command:

python eval_voc.py --device 0 --test --resume path/to/weight.pth

To evaluate on the ADE20K dataset, execute the following command:

python eval_ade.py --device 0 --test --resume path/to/weight.pth

To evaluate on the CityScapes dataset, execute the following command:

python eval_city.py --device 0 --test --resume path/to/weight.pth

@article{chen2025replay,
  title={Replay Without Saving: Prototype Derivation and Distribution Rebalance for Class-Incremental Semantic Segmentation},
  author={Chen, Jinpeng and Cong, Runmin and Yuxuan, Luo and Ip, Horace and Kwong, Sam},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
  year={2025},
  volume={47},
  number={6},
  pages={4699-4716}
}

• This code is based on DKD ([2022-NeurIPS] Decomposed Knowledge Distillation for Class-Incremental Semantic Segmentation).
• This template is borrowed from pytorch-template.