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

We design a family of image classification architectures that optimize thetrade-off between accuracy and efficiency in a high-speed regime. Our workexploits recent findings in attention-based architectures, which arecompetitive on highly parallel processing hardware. We revisit principles fromthe extensive literature on convolutional neural networks to apply them totransformers, in particular activation maps with decreasing resolutions. Wealso introduce the attention bias, a new way to integrate positionalinformation in vision transformers. As a result, we propose LeVIT: a hybridneural network for fast inference image classification. We consider differentmeasures of efficiency on different hardware platforms, so as to best reflect awide range of application scenarios. Our extensive experiments empiricallyvalidate our technical choices and show they are suitable to mostarchitectures. Overall, LeViT significantly outperforms existing convnets andvision transformers with respect to the speed/accuracy tradeoff. For example,at 80% ImageNet top-1 accuracy, LeViT is 5 times faster than EfficientNet onCPU. We release the code at https://github.com/facebookresearch/LeViT

LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference