Awesome

UVO_Challenge

High quality class-agnostic image/video segmentation framework.

Update

Please kindly refer to this repo for the detection and segmentation code. This is exactly the same model we use during the UVO challenge.

Visualization

Our method can still achieve decent results on blur/occluded images/videos.

Video

Team Alpes_runner Solutions

This is an official repo for our UVO Challenge solutions for Image/Video-based open-world segmentation. Our team "Alpes_runner" achieved the best performance on both Image/Video-based benchmarks. More details about the workshop can be found here.

Technical Reports

For Track 1: paper
For Track 2: paper

Models

Detection

Model	Pretrained datasets	Finetuned datasets	links
UVO_Detector	COCO	-	config/weights
UVO_Detector	COCO	UVO train+val	config/weights

Segmentation

Model	Pretrained datasets	Finetuned datasets	links
UVO_Segementor	COCO	-	weights
UVO_Segmentor	COCO, PASCAL, OpenImage	-	config/weights
UVO_Segmentor	COCO, PASCAL, OpenImage	UVO train+val	config/weights

Citation

If you find this project useful in your research, please consider cite:

@article{du20211st,
  title={1st Place Solution for the UVO Challenge on Image-based Open-World Segmentation 2021},
  author={Du, Yuming and Guo, Wen and Xiao, Yang and Lepetit, Vincent},
  journal={arXiv preprint arXiv:2110.10239},
  year={2021}
}

@article{du20211st,
  title={1st Place Solution for the UVO Challenge on Video-based Open-World Segmentation 2021},
  author={Du, Yuming and Guo, Wen and Xiao, Yang and Lepetit, Vincent},
  journal={arXiv preprint arXiv:2110.11661},
  year={2021}
}

Contact

Feel free to open a new issue(Prefered, as everybody can see your issue) or contact me if you have any questions.