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

The Mixup scheme suggests mixing a pair of samples to create an augmentedtraining sample and has gained considerable attention recently for improvingthe generalizability of neural networks. A straightforward and widely usedextension of Mixup is to combine with regional dropout-like methods: removingrandom patches from a sample and replacing it with the features from anothersample. Albeit their simplicity and effectiveness, these methods are prone tocreate harmful samples due to their randomness. To address this issue, 'maximumsaliency' strategies were recently proposed: they select only the mostinformative features to prevent such a phenomenon. However, they now sufferfrom lack of sample diversification as they always deterministically selectregions with maximum saliency, injecting bias into the augmented data. In thispaper, we present, a novel, yet simple Mixup-variant that captures the best ofboth worlds. Our idea is two-fold. By stochastically sampling the features and'grafting' them onto another sample, our method effectively generates diverseyet meaningful samples. Its second ingredient is to produce the label of thegrafted sample by mixing the labels in a saliency-calibrated fashion, whichrectifies supervision misguidance introduced by the random sampling procedure.Our experiments under CIFAR, Tiny-ImageNet, and ImageNet datasets show that ourscheme outperforms the current state-of-the-art augmentation strategies notonly in terms of classification accuracy, but is also superior in coping understress conditions such as data corruption and object occlusion.

Saliency Grafting: Innocuous Attribution-Guided Mixup with Calibrated Label Mixing