Abstract: Low-shot image classification, where training images are limited or inaccessible, has benefited from recent progress on pretrained vision-language (VL) models with strong generalizability. e ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback