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

Advanced data augmentation strategies have widely been studied to improve thegeneralization ability of deep learning models. Regional dropout is one of thepopular solutions that guides the model to focus on less discriminative partsby randomly removing image regions, resulting in improved regularization.However, such information removal is undesirable. On the other hand, recentstrategies suggest to randomly cut and mix patches and their labels amongtraining images, to enjoy the advantages of regional dropout without having anypointless pixel in the augmented images. We argue that such random selectionstrategies of the patches may not necessarily represent sufficient informationabout the corresponding object and thereby mixing the labels according to thatuninformative patch enables the model to learn unexpected featurerepresentation. Therefore, we propose SaliencyMix that carefully selects arepresentative image patch with the help of a saliency map and mixes thisindicative patch with the target image, thus leading the model to learn moreappropriate feature representation. SaliencyMix achieves the best known top-1error of 21.26% and 20.09% for ResNet-50 and ResNet-101 architectures onImageNet classification, respectively, and also improves the model robustnessagainst adversarial perturbations. Furthermore, models that are trained withSaliencyMix help to improve the object detection performance. Source code isavailable at https://github.com/SaliencyMix/SaliencyMix.

SaliencyMix: A Saliency Guided Data Augmentation Strategy for Better Regularization