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

Human visual recognition is a sparse process, where only a few salient visualcues are attended to rather than traversing every detail uniformly. However,most current vision networks follow a dense paradigm, processing every singlevisual unit (e.g,, pixel or patch) in a uniform manner. In this paper, wechallenge this dense paradigm and present a new method, coined SparseFormer, toimitate human's sparse visual recognition in an end-to-end manner. SparseFormerlearns to represent images using a highly limited number of tokens (down to 49)in the latent space with sparse feature sampling procedure instead ofprocessing dense units in the original pixel space. Therefore, SparseFormercircumvents most of dense operations on the image space and has much lowercomputational costs. Experiments on the ImageNet classification benchmarkdataset show that SparseFormer achieves performance on par with canonical orwell-established models while offering better accuracy-throughput tradeoff.Moreover, the design of our network can be easily extended to the videoclassification with promising performance at lower computational costs. We hopethat our work can provide an alternative way for visual modeling and inspirefurther research on sparse neural architectures. The code will be publiclyavailable at https://github.com/showlab/sparseformer

SparseFormer: Sparse Visual Recognition via Limited Latent Tokens