Reproduce “Cost Aggregation with 4D Convolutional Swin Transformer for Few-Shot Segmentation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Sunghwan Hong, Seokju Cho, Jisu Nam, Stephen Lin, Seungryong Kim

VAT presents few-shot segmentation as pixel matching between labeled support images and query images. Its 4D convolutional Swin Transformer combines overlapping processing with shifted-window attention, while coarse-to-fine guidance and query appearance help refine masks. On FSS-1000, with ResNet-50 in the 1-shot setting, it reports 90.1% mIoU versus HSNet’s 85.5%. This is benchmark evidence under the protocols, not proof of universal performance.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.