Skip to content

helen1c/DEARLi

Repository files navigation

DEARLi

**Official implementation of the ICCV 2025 Findings Poster & Oral: **
"DEARLi: Decoupled Enhancement of Recognition and Localization for Semi-Supervised Panoptic Segmentation"

DEARLi Overview

Getting Started

For detailed setup instructions -- including installation, dataset preparation, checkpoints, training, and evaluation -- please see:
readmes/GETTING_STARTED.md


TODO

  • Upload COCO-Objects dataset generation script and panoptic labels
  • Upload SAM-generated pseudolabels and add instructions to generate them

Acknowledgments

This project builds upon several excellent open-source frameworks.
We thank the respective authors for making their code publicly available:


Citation

If you use DEARLi in your research, please cite the following:

@article{martinovic2025dearli,
  title={DEARLi: Decoupled Enhancement of Recognition and Localization for Semi-Supervised Panoptic Segmentation},
  author={Martinović, Ivan and Šarić, Josip and Oršić, Marin and Kristan, Matej and Šegvić, Siniša},
  journal={arXiv preprint arXiv:2507.10118},
  year={2025}
}

About

This repo contains the code for our ICCV'25 Findings paper "DEARLi: Decoupled Enhancement of Recognition and Localization for Semi-supervised Panoptic Segmentation"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors