English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Pre-trained vision-language models have inspired much research on few-shotlearning. However, with only a few training images, there exist two crucialproblems: (1) the visual feature distributions are easily distracted byclass-irrelevant information in images, and (2) the alignment between thevisual and language feature distributions is difficult. To deal with thedistraction problem, we propose a Selective Attack module, which consists oftrainable adapters that generate spatial attention maps of images to guide theattacks on class-irrelevant image areas. By messing up these areas, thecritical features are captured and the visual distributions of image featuresare calibrated. To better align the visual and language feature distributionsthat describe the same object class, we propose a cross-modal distributionalignment module, in which we introduce a vision-language prototype for eachclass to align the distributions, and adopt the Earth Mover's Distance (EMD) tooptimize the prototypes. For efficient computation, the upper bound of EMD isderived. In addition, we propose an augmentation strategy to increase thediversity of the images and the text prompts, which can reduce overfitting tothe few-shot training images. Extensive experiments on 11 datasets demonstratethat our method consistently outperforms prior arts in few-shot learning. Theimplementation code will be available at https://github.com/bhrqw/SADA.

Few-Shot Learning with Visual Distribution Calibration and Cross-Modal Distribution Alignment