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

Deep convolutional neural networks achieve remarkable visual recognitionperformance, at the cost of high computational complexity. In this paper, wehave a new design of efficient convolutional layers based on three schemes. The3D convolution operation in a convolutional layer can be considered asperforming spatial convolution in each channel and linear projection acrosschannels simultaneously. By unravelling them and arranging the spatialconvolution sequentially, the proposed layer is composed of a singleintra-channel convolution, of which the computation is negligible, and a linearchannel projection. A topological subdivisioning is adopted to reduce theconnection between the input channels and output channels. Additionally, wealso introduce a spatial "bottleneck" structure that utilizes aconvolution-projection-deconvolution pipeline to take advantage of thecorrelation between adjacent pixels in the input. Our experiments demonstratethat the proposed layers remarkably outperform the standard convolutionallayers with regard to accuracy/complexity ratio. Our models achieve similaraccuracy to VGG, ResNet-50, ResNet-101 while requiring 42, 4.5, 6.5 times lesscomputation respectively.

Design of Efficient Convolutional Layers using Single Intra-channel Convolution, Topological Subdivisioning and Spatial "Bottleneck" Structure