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

Standard convolutional neural networks(CNNs) require consistent imageresolutions in both training and testing phase. However, in practice, testingwith smaller image sizes is necessary for fast inference. We show thattrivially evaluating low-resolution images on networks trained withhigh-resolution images results in a catastrophic accuracy drop in standard CNNarchitectures. We propose a novel training regime called Scale calibratedTraining(SCT) which allows networks to learn from various scales of inputsimultaneously. By taking advantages of SCT, single network can provide decentaccuracy at test time in response to multiple test scales. In our analysis, wesurprisingly find that vanilla batch normalization can lead to sub-optimalperformance in SCT. Therefore, a novel normalization scheme calledScale-Specific Batch Normalization is equipped to SCT in replacement of batchnormalization. Experiment results show that SCT improves accuracy of singleResnet-50 on ImageNet by 1.7% and 11.5% accuracy when testing on image sizes of224 and 128 respectively.

Scale Calibrated Training: Improving Generalization of Deep Networks via Scale-Specific Normalization