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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Binary neural network (BNN) is an extreme quantization version ofconvolutional neural networks (CNNs) with all features and weights mapped tojust 1-bit. Although BNN saves a lot of memory and computation demand to makeCNN applicable on edge or mobile devices, BNN suffers the drop of networkperformance due to the reduced representation capability after binarization. Inthis paper, we propose a new replaceable and easy-to-use convolution moduleRepConv, which enhances feature maps through replicating input or output alongchannel dimension by $\beta$ times without extra cost on the number ofparameters and convolutional computation. We also define a set of RepTran rulesto use RepConv throughout BNN modules like binary convolution, fully connectedlayer and batch normalization. Experiments demonstrate that after the RepTrantransformation, a set of highly cited BNNs have achieved universally betterperformance than the original BNN versions. For example, the Top-1 accuracy ofRep-ReCU-ResNet-20, i.e., a RepBconv enhanced ReCU-ResNet-20, reaches 88.97% onCIFAR-10, which is 1.47% higher than that of the original network. AndRep-AdamBNN-ReActNet-A achieves 71.342% Top-1 accuracy on ImageNet, a freshstate-of-the-art result of BNNs. Code and models are availableat:https://github.com/imfinethanks/Rep_AdamBNN.

RepBNN: towards a precise Binary Neural Network with Enhanced Feature Map via Repeating