Efficient Masked Attention Transformer for Few-Shot Classification and Segmentation
Loading...
Date
2025-08-26
Type
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Description
Few-shot classification and segmentation (FS-CS) focuses on jointly performing multi-label classification and multi-class segmentation using few annotated examples. Although the current state of the art (SOTA) achieves high accuracy in both tasks, it struggles with small objects. To overcome this, we propose the Efficient Masked Attention Transformer (EMAT), which improves classification and segmentation accuracy, especially for small objects. EMAT introduces three modifications: a novel memory-efficient masked attention mechanism, a learnable downscaling strategy, and parameter-efficiency enhancements. EMAT outperforms all FS-CS methods on the PASCAL-5i and COCO-20i datasets, using at least four times fewer trainable parameters. Moreover, as the current FS-CS evaluation setting discards available annotations, despite their costly collection, we introduce two novel evaluation settings that consider these annotations to better reflect practical scenarios.
Citation
Endorsement
Related Resources
Is Supplement To
https://arxiv.org/abs/2507.23642Faculty
Collections
License
Except where otherwise noted, this license is described as Apache License 2.0