MSI: Maximize Support-Set Information for Few-Shot Segmentation (ICCV 2023)

The paper is on [arXiv].

VAT-MSI Pretrained models Link

• Pascal-5 Benchmark with ResNet50
• Pascal-5 Benchmark with ResNet101
• COCO-20 Benchmark with ResNet50
• COCO-20 Benchmark with ResNet101
• FSS-1000 Benchmark with ResNet50
• FSS-1000 Benchmark with ResNet101

Performance

Visualization

References

Our work is based on these models. (HSNet, VAT, ASNet, and HM)

• HSNet : Hypercorrelation Squeeze for Few-Shot Segmentation, ICCV 2021
• VAT : Cost Aggregation with 4D Convolutional Swin Transformer for Few-Shot Segmentation, ECCV 2022
• HM : Hybrid Masking for Few-Shot Segmentation, ECCV 2022
• ASNet : Integrative Few-Shot Learning for Classification and Segmentation, CVPR 2022

Thank you very much.

BibTeX

If you find this research useful, please consider citing:

@misc{moon2023msi,
      title={MSI: Maximize Support-Set Information for Few-Shot Segmentation}, 
      author={Seonghyeon Moon and Samuel S. Sohn and Honglu Zhou and Sejong Yoon and Vladimir Pavlovic and Muhammad Haris Khan and Mubbasir Kapadia},
      year={2023},
      eprint={2212.04673},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}