lixiang007666 / wscl

The official PyTorch implementation of our paper (Weak-to-Strong Consistency Learning for Semisupervised Image Segmentation) Accepted by TGRS2023

WSCL

This is the official PyTorch implementation of our paper:

Weak-to-Strong Consistency Learning for Semisupervised Image Segmentation
Xiaoqiang Lu, Licheng Jiao, Lingling Li, Fang Liu, Xu Liu, Shuyuan Yang, Zhixi Feng, Puhua Chen
Accepted by IEEE Transactions on Geoscience and Remote Sensing (TGRS) 2023

Getting Started

Install

conda create -n WSCL python=3.7
pip install -r requirements.txt

Data Preparation

Pre-trained Model

mkdir pretrained
cd pretrained
wget https://download.pytorch.org/models/resnet101-63fe2227.pth
mv resnet101-63fe2227.pth resnet101.pth
cd ..

or download via the following link ResNet-101

Dataset

We have processed the original dataset as mentioned in the paper. You can access the processed dataset directly via the following link.

DFC22 | iSAID | MER | MSL | Vaihingen | GID-15 |

File Organization

├── ./pretrained
    └── resnet101.pth
    
├── [Your Dataset Path]
    ├── images
    └── labels

Results

Dataset	Partition	Method	mIoU	Dataset	Partition	Method	mIoU
DFC22	1-8	baseline	26.97	iSAID	100	baseline	39.91
		LSST	30.94			LSST	46.94
		Ours	38.00			Ours	59.60
	1-4	baseline	32.67		300	baseline	60.47
		LSST	36.40			LSST	63.40
		Ours	38.92			Ours	72.33
MER	1-8	baseline	43.63	MSL	1-8	baseline	50.17
		LSST	49.68			LSST	54.72
		Ours	51.88			Ours	58.23
	1-4	baseline	48.19		1-4	baseline	50.26
		LSST	51.31			LSST	56.22
		Ours	54.85			Ours	59.91
Vaihingen	1-8	baseline	53.30	GID-15	1-8	baseline	61.86
		LSST	64.09			LSST	66.38
		Ours	66.65			Ours	71.42
	1-4	baseline	59.30		1-4	baseline	67.90
		LSST	65.34			LSST	71.28
		Ours	68.81			Ours	74.88

Training and Testing

Training

Change DATASET, SPLIT, and DATASET_PATH as you want in train.py, then run:

CUDA_VISIBLE_DEVICES=0,1 python train.py

Testing

Change WEIGHTS, and DATASET_PATH as you want in test.py, then run:

CUDA_VISIBLE_DEVICES=0,1 python test.py

Citation

If you find this project useful, please consider citing:

@ARTICLE{10114409,
  author={Lu, Xiaoqiang and Jiao, Licheng and Li, Lingling and Liu, Fang and Liu, Xu and Yang, Shuyuan and Feng, Zhixi and Chen, Puhua},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={Weak-to-Strong Consistency Learning for Semisupervised Image Segmentation}, 
  year={2023},
  volume={61},
  number={},
  pages={1-15},
  doi={10.1109/TGRS.2023.3272552}} 

We have other work on semi-supervised remote sensing image segmentation:

• [IEEE TGRS 2022] LSST