Hunting Sparsity: Density-Guided Contrastive Learning for Semi-Supervised Semantic Segmentation
Xiaoyang Wang, Bingfeng Zhang, Limin Yu, and Jimin Xiao.
In CVPR 2023.
Abstract: Inspired by density-based unsupervised clustering, we propose to leverage feature density to locate sparse regions within feature clusters defined by label and pseudo labels. The hypothesis is that lower-density features tend to be under-trained compared with those densely gathered. Therefore, we propose to apply regularization on the structure of the cluster by tackling the sparsity to increase intra-class compactness in feature space. With this goal, we present a Density-Guided Contrastive Learning (DGCL) strategy to push anchor features in sparse regions toward cluster centers approximated by high-density positive keys. The heart of our method is to estimate feature density which is defined as neighbor compactness. We design a multi-scale density estimation module to obtain the density from multiple nearest-neighbor graphs for robust density modeling. Moreover, a unified training framework is proposed to combine label-guided self-training and densityguided geometry regularization to form complementary supervision on unlabeled data.
cd DGCL
conda create -n dgcl python=3.10
conda activate dgcl
pip install -r requirements.txtDownload pretrained wegiths ResNet-101
├── DGCL/
└── resnet101.pth
├── Path_to_Pascal
├── JPEGImages
└── SegmentationClassAug
├── Path_to_Cityscapes
├── leftImg8bit
└── gtFine
Navigate into experiments/pascal/732 and modify config.yaml and train.sh.
sh train.sh <num_gpu> <port>
@inproceedings{wang2023dgcl,
title= {Hunting Sparsity: Density-Guided Contrastive Learning for Semi-Supervised Semantic Segmentation},
author={Wang, Xiaoyang and Zhang, Bingfeng and Yu, Limin and Xiao, Jimin},
booktitle={CVPR},
year={2023},
}This project borrows codes from U2PL and ReCo. Thanks for their great work!
For questions, please contact: wangxy@liverpool.ac.uk