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

Aiming towards human-level generalization, there is a need to exploreadaptable representation learning methods with greater transferability. Mostexisting approaches independently address task-transferability and cross-domainadaptation, resulting in limited generalization. In this paper, we proposeUM-Adapt - a unified framework to effectively perform unsupervised domainadaptation for spatially-structured prediction tasks, simultaneouslymaintaining a balanced performance across individual tasks in a multi-tasksetting. To realize this, we propose two novel regularization strategies; a)Contour-based content regularization (CCR) and b) exploitation of inter-taskcoherency using a cross-task distillation module. Furthermore, avoiding aconventional ad-hoc domain discriminator, we re-utilize the cross-taskdistillation loss as output of an energy function to adversarially minimize theinput domain discrepancy. Through extensive experiments, we demonstratesuperior generalizability of the learned representations simultaneously formultiple tasks under domain-shifts from synthetic to natural environments.UM-Adapt yields state-of-the-art transfer learning results on ImageNetclassification and comparable performance on PASCAL VOC 2007 detection task,even with a smaller backbone-net. Moreover, the resulting semi-supervisedframework outperforms the current fully-supervised multi-task learningstate-of-the-art on both NYUD and Cityscapes dataset.

UM-Adapt: Unsupervised Multi-Task Adaptation Using Adversarial Cross-Task Distillation