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

Class-agnostic counting (CAC) aims to count objects of interest from a queryimage given few exemplars. This task is typically addressed by extracting thefeatures of query image and exemplars respectively with (un)shared featureextractors and by matching their feature similarity, leading to anextract-\textit{then}-match paradigm. In this work, we show that CAC can besimplified in an extract-\textit{and}-match manner, particularly using apretrained and plain vision transformer (ViT) where feature extraction andsimilarity matching are executed simultaneously within the self-attention. Wereveal the rationale of such simplification from a decoupled view of theself-attention and point out that the simplification is only made possible ifthe query and exemplar tokens are concatenated as input. The resulting model,termed CACViT, simplifies the CAC pipeline and unifies the feature spacesbetween the query image and exemplars. In addition, we find CACViT naturallyencodes background information within self-attention, which helps reducebackground disturbance. Further, to compensate the loss of the scale and theorder-of-magnitude information due to resizing and normalization in ViT, wepresent two effective strategies for scale and magnitude embedding. Extensiveexperiments on the FSC147 and the CARPK datasets show that CACViT significantlyoutperforms state-of-the-art CAC approaches in both effectiveness (23.60% errorreduction) and generalization, which suggests CACViT provides a concise andstrong baseline for CAC. Code will be available.

Vision Transformer Off-the-Shelf: A Surprising Baseline for Few-Shot Class-Agnostic Counting