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

Existing works often focus on reducing the architecture redundancy foraccelerating image classification but ignore the spatial redundancy of theinput image. This paper proposes an efficient image classification pipeline tosolve this problem. We first pinpoint task-aware regions over the input imageby a lightweight patch proposal network called AnchorNet. We then feed theselocalized semantic patches with much smaller spatial redundancy into a generalclassification network. Unlike the popular design of deep CNN, we aim tocarefully design the Receptive Field of AnchorNet without intermediateconvolutional paddings. This ensures the exact mapping from a high-levelspatial location to the specific input image patch. The contribution of eachpatch is interpretable. Moreover, AnchorNet is compatible with any downstreamarchitecture. Experimental results on ImageNet show that our method outperformsSOTA dynamic inference methods with fewer inference costs. Our code isavailable at https://github.com/winycg/AnchorNet.

Localizing Semantic Patches for Accelerating Image Classification