The Loupe: A Plug-and-Play Attention Module for Amplifying Discriminative Features in Vision Transformers

In three linesThe Loupe is a lightweight spatial gating module for hierarchical Vision Transformers designed for fine-grained visual classification. Inserted at an intermediate feature stage, it predicts a single-channel spatial mask via a small CNN and reweights activations. On CUB-200-2011, it improves Swin-Base from 88.36% to 91.72% and Swin-Tiny from 85.14% to 88.61% with <0.1% additional parameters.

Summary generated by Claude — human-verified