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

For video and volumetric data understanding, 3D convolution layers are widelyused in deep learning, however, at the cost of increasing computation andtraining time. Recent works seek to replace the 3D convolution layer withconvolution blocks, e.g. structured combinations of 2D and 1D convolutionlayers. In this paper, we propose a novel convolution block, Parallel Separable3D Convolution (PmSCn), which applies m parallel streams of n 2D and one 1Dconvolution layers along different dimensions. We first mathematically justifythe need of parallel streams (Pm) to replace a single 3D convolution layerthrough tensor decomposition. Then we jointly replace consecutive 3Dconvolution layers, common in modern network architectures, with the multiple2D convolution layers (Cn). Lastly, we empirically show that PmSCn isapplicable to different backbone architectures, such as ResNet, DenseNet, andUNet, for different applications, such as video action recognition, MRI brainsegmentation, and electron microscopy segmentation. In all three applications,we replace the 3D convolution layers in state-of-the art models with PmSCn andachieve around 14% improvement in test performance and 40% reduction in modelsize and on average.

Parallel Separable 3D Convolution for Video and Volumetric Data Understanding