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

Vision transformers (ViTs) are usually considered to be less light-weightthan convolutional neural networks (CNNs) due to the lack of inductive bias.Recent works thus resort to convolutions as a plug-and-play module and embedthem in various ViT counterparts. In this paper, we argue that theconvolutional kernels perform information aggregation to connect all tokens;however, they would be actually unnecessary for light-weight ViTs if thisexplicit aggregation could function in a more homogeneous way. Inspired bythis, we present LightViT as a new family of light-weight ViTs to achievebetter accuracy-efficiency balance upon the pure transformer blocks withoutconvolution. Concretely, we introduce a global yet efficient aggregation schemeinto both self-attention and feed-forward network (FFN) of ViTs, whereadditional learnable tokens are introduced to capture global dependencies; andbi-dimensional channel and spatial attentions are imposed over tokenembeddings. Experiments show that our model achieves significant improvementson image classification, object detection, and semantic segmentation tasks. Forexample, our LightViT-T achieves 78.7% accuracy on ImageNet with only 0.7GFLOPs, outperforming PVTv2-B0 by 8.2% while 11% faster on GPU. Code isavailable at https://github.com/hunto/LightViT.

LightViT: Towards Light-Weight Convolution-Free Vision Transformers