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 3-dimensional (3D) Convolutional Network (ConvNet) has shown promisingperformance on video recognition tasks because of its powerful spatio-temporalinformation fusion ability. However, the extremely intensive requirements onmemory access and computing power prohibit it from being used inresource-constrained scenarios, such as portable and edge devices. So in thispaper, we first propose a two-stage Fully Separable Block (FSB) tosignificantly compress the model sizes of 3D ConvNets. Then a featureenhancement approach named Temporal Residual Gradient (TRG) is developed toimprove the performance of compressed model on video tasks, which provideshigher accuracy, faster convergency and better robustness. Moreover, in orderto further decrease the computing workload, we propose a hybrid Fast Algorithm(hFA) to drastically reduce the computation complexity of convolutions. Thesemethods are effectively combined to design a light-weight and efficient ConvNetfor video recognition tasks. Experiments on the popular dataset report 2.3xcompression rate, 3.6x workload reduction, and 6.3% top-1 accuracy gain, overthe state-of-the-art SlowFast model, which is already a highly compact model.The proposed methods also show good adaptability on traditional 3D ConvNet,demonstrating 7.4x more compact model, 11.0x less workload, and 3.0% higheraccuracy

Design Light-weight 3D Convolutional Networks for Video Recognition Temporal Residual, Fully Separable Block, and Fast Algorithm