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

In fine-grained image recognition (FGIR), the localization and amplificationof region attention is an important factor, which has been explored a lot byconvolutional neural networks (CNNs) based approaches. The recently developedvision transformer (ViT) has achieved promising results on computer visiontasks. Compared with CNNs, Image sequentialization is a brand new manner.However, ViT is limited in its receptive field size and thus lacks localattention like CNNs due to the fixed size of its patches, and is unable togenerate multi-scale features to learn discriminative region attention. Tofacilitate the learning of discriminative region attention without box/partannotations, we use the strength of the attention weights to measure theimportance of the patch tokens corresponding to the raw images. We propose therecurrent attention multi-scale transformer (RAMS-Trans), which uses thetransformer's self-attention to recursively learn discriminative regionattention in a multi-scale manner. Specifically, at the core of our approachlies the dynamic patch proposal module (DPPM) guided region amplification tocomplete the integration of multi-scale image patches. The DPPM starts with thefull-size image patches and iteratively scales up the region attention togenerate new patches from global to local by the intensity of the attentionweights generated at each scale as an indicator. Our approach requires only theattention weights that come with ViT itself and can be easily trainedend-to-end. Extensive experiments demonstrate that RAMS-Trans performs betterthan concurrent works, in addition to efficient CNN models, achievingstate-of-the-art results on three benchmark datasets.

RAMS-Trans: Recurrent Attention Multi-scale Transformer forFine-grained Image Recognition