cv516Buaa/STCL-DRNet

STCL-DRNet

This repo is the implementation of "Self-Training and Curriculum Learning Guided Dynamic Refined Network for Remote Sensing Class-Incremental Semantic Segmentation"

Dataset Preparation

We select DeepGlobe, iSAID as benchmark datasets and create train, val, test list for researchers to follow.

In the following, we provide the detailed commands for dataset preparation.

DeepGlobe

  download the DeepGlobe dataset and unzip and move to the data/DeepGlobe2018 folder
  python data/rgb2label.py to generate the one-channel label and the data folder

iSAID

  we use the mmsegmentation/tools/dataset_converters/iSAID.py to crop the picture to 512*512 overlaping 384 
  and generate the one channel labels.

Datasets structures

data/
    --- DeepGlobe2018
        --- land_train/
        --- onechannel_label/
    --- iSAID
        --- img_dir
            --- train
            --- val
            --- test
        --- ann_dir
            --- train
            --- val
            --- test

Install

 ```
 pip install -r requirements.txt
 ``` 

Training

1. Class-Incremental Segmentation on DeepGlobe

   cd STCL-DRNet/scripts/train
   
   sh DeepGlobe_3-3.sh # 3-3 incremental learning
   sh DeepGlobe_2-2.sh # 2-2 incremental learning
   sh DeepGlobe_1-1.sh # 1-1 incremental learning
   

2. Class-Incremental Segmentation on iSAID:

   cd STCL-DRNet/scripts/train
   
   sh iSAID_14-1.sh # 14-1 incremental learning
   sh iSAID_10-5.sh # 10-5 incremental learning
   sh iSAID_10-1.sh # 10-1 incremental learning
   

Testing

Trained with the above commands, you can get a trained model to test the performance of your model.

1. test on DeepGlobe

    cd STCL-DRNet/scripts/test
    
    sh test_DeepGlobe_3-3.sh # 3-3 incremental learning
    sh test_DeepGlobe_2-2.sh # 2-2 incremental learning
    sh test_DeepGlobe_1-1.sh # 1-1 incremental learning
   

2. test on iSAID

   cd STCL-DRNet/scripts/train
    
    sh test_iSAID_14-1.sh # 14-1 incremental learning
    sh test_iSAID_10-5.sh # 10-5 incremental learning
    sh test_iSAID_10-1.sh # 10-1 incremental learning
   

Results

> >

If you have any question, please discuss with me by sending email to lyushuchang@buaa.edu.cn.

References

Many thanks to their excellent works SSUL

STCL-DRNet