Home

Awesome

[CVPR 2022] Unbiased Subclass Regularization for Semi-Supervised Semantic Segmentation

Updates

Paper

Unbiased Subclass Regularization for Semi-Supervised Semantic Segmentation

Dayan Guan, Jiaxing Huang, Xiao Aoran, Shijian Lu
School of Computer Science and Engineering, Nanyang Technological University, Singapore

Abstract

Semi-supervised semantic segmentation learns from small amounts of labelled images and large amounts of unlabelled images, which has witnessed impressive progress with the recent advance of deep neural networks. However, it often suffers from severe class-bias problem while exploring the unlabelled images, largely due to the clear pixel-wise class imbalance in the labelled images. This paper presents an unbiased subclass regularization network (USRN) that alleviates the class imbalance issue by learning class-unbiased segmentation from balanced subclass distributions. We build the balanced subclass distributions by clustering pixels of each original class into multiple subclasses of similar sizes, which provide class-balanced pseudo supervision to regularize the class-biased segmentation. In addition, we design an entropy-based gate mechanism to coordinate learning between the original classes and the clustered subclasses which facilitates subclass regularization effectively by suppressing unconfident subclass predictions. Extensive experiments over multiple public benchmarks show that USRN achieves superior performance as compared with the state-of-the-art.

Preparation

  1. Environment:
git clone https://github.com/Dayan-Guan/USRN.git
cd USRN
sh init.sh
  1. dataset:
USRN/dataset/voc/VOCdevkit/VOC2012                          % PASCAL VOC 2012 dataset root
USRN/dataset/voc/VOCdevkit/VOC2012/JPEGImages               % Images
USRN/dataset/voc/VOCdevkit/VOC2012/SegmentationClass        % Annotations
USRN/dataset/voc/VOCdevkit/VOC2012/SegmentationClassAug     % Extra annotations
USRN/dataset/cityscapes/                       % cityscapes dataset root
USRN/dataset/cityscapes/leftImg8bit_sequence   % leftImg8bit_trainvaltest
USRN/dataset/cityscapes/images                 % cp ../leftImg8bit_sequence/train/*/* ./images/train/
USRN/dataset/cityscapes/gtFine                 % gtFine_trainvaltest
USRN/dataset/cityscapes/segmentation           % cp ../gtFine_trainvaltest/train/*/* ./segmentation/train/
  1. Pre-trained models: Download pre-trained models and put in USRN/pretrained

Evaluation and Visualization using Pretrained Models

python3 main.py --test True --resume pretrained/best_model_voc_1over32_baseline.pth --config configs/voc_1over32_baseline.json
python3 main.py --test True --resume pretrained/best_model_voc_1over32_usrn.pth --config configs/voc_1over32_usrn.json

Training and Testing

sh main.sh voc_1over32

Acknowledgements

This codebase is heavily borrowed from CAC.

Contact

If you have any questions, please contact: dayan.guan@outlook.com

Citation

If you find this code useful for your research, please cite our paper:

@inproceedings{guan2022unbiased,
  title={Unbiased Subclass Regularization for Semi-Supervised Semantic Segmentation},
  author={Guan, Dayan and Huang, Jiaxing and Xiao, Aoran and Lu, Shijian},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={9968--9978},
  year={2022}
}