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

Deploying deep visual models can lead to performance drops due to thediscrepancies between source and target distributions. Several approachesleverage labeled source data to estimate target domain accuracy, but accessinglabeled source data is often prohibitively difficult due to dataconfidentiality or resource limitations on serving devices. Our work proposes anew framework to estimate model accuracy on unlabeled target data withoutaccess to source data. We investigate the feasibility of using pseudo-labelsfor accuracy estimation and evolve this idea into adopting recent advances insource-free domain adaptation algorithms. Our approach measures thedisagreement rate between the source hypothesis and the target pseudo-labelingfunction, adapted from the source hypothesis. We mitigate the impact oferroneous pseudo-labels that may arise due to a high ideal joint hypothesisrisk by employing adaptive adversarial perturbation on the input of the targetmodel. Our proposed source-free framework effectively addresses the challengingdistribution shift scenarios and outperforms existing methods requiring sourcedata and labels for training.

Unsupervised Accuracy Estimation of Deep Visual Models using Domain-Adaptive Adversarial Perturbation without Source Samples