Awesome
U2Seg_eff
This is a sub-part of U2Seg, which reproduce the Efficient Learning part in the main paper. See README.md for details.
<p align="center"> <img src='./teaser_img.jpg' align="center" > </p>Unsupervised Universal Image Segmentation
Dantong Niu*, Xudong Wang*, Xinyang Han*, Long Lian, Roei Herzig, Trevor Darrell
Berkeley AI Research, UC Berkeley
CVPR 2024
project page | arxiv | bibtex
<a href="https://paperswithcode.com/sota/unsupervised-panoptic-segmentation-on-coco?p=unsupervised-universal-image-segmentation"> <img src="https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/unsupervised-universal-image-segmentation/unsupervised-panoptic-segmentation-on-coco"> </a> <br> <a href="https://paperswithcode.com/sota/unsupervised-zero-shot-panoptic-segmentation?p=unsupervised-universal-image-segmentation"> <img src="https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/unsupervised-universal-image-segmentation/unsupervised-zero-shot-panoptic-segmentation"> </a> <br> <a href="https://paperswithcode.com/sota/unsupervised-zero-shot-instance-segmentation?p=unsupervised-universal-image-segmentation"> <img src="https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/unsupervised-universal-image-segmentation/unsupervised-zero-shot-instance-segmentation"> </a> <br> <a href="https://paperswithcode.com/sota/unsupervised-semantic-segmentation-on-coco-7?p=unsupervised-universal-image-segmentation"> <img src="https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/unsupervised-universal-image-segmentation/unsupervised-semantic-segmentation-on-coco-7"> </a> <!-- ## Features - U2Seg is the first universal unsupervised image segmentation model that can tackle unsupervised semantic-aware instance, semantic and panoptic segmentation tasks using a unified framework. - U2Seg can learn unsupervised object detectors and instance segmentors solely on ImageNet-1K. - U2Seg exhibits strong robustness to domain shifts when evaluated on 11 different benchmarks across domains like natural images, video frames, paintings, sketches, etc. - U2Seg can serve as a pretrained model for fully/semi-supervised detection and segmentation tasks. -->Citation
If you find our work inspiring or use our codebase in your research, please consider giving a star ⭐ and a citation.
@misc{niu2023unsupervised,
title={Unsupervised Universal Image Segmentation},
author={Dantong Niu and Xudong Wang and Xinyang Han and Long Lian and Roei Herzig and Trevor Darrell},
year={2023},
eprint={2312.17243},
archivePrefix={arXiv},
primaryClass={cs.CV}
}