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

Large-scale pre-trained models have achieved remarkable success in variouscomputer vision tasks. A standard approach to leverage these models is tofine-tune all model parameters for downstream tasks, which poses challenges interms of computational and storage costs. Recently, inspired by NaturalLanguage Processing (NLP), parameter-efficient transfer learning has beensuccessfully applied to vision tasks. However, most existing techniquesprimarily focus on single-task adaptation, and despite limited research onmulti-task adaptation, these methods often exhibit suboptimal training andinference efficiency. In this paper, we first propose an once-for-all VisionMulti-Task Adapter (VMT-Adapter), which strikes approximately O(1) training andinference efficiency w.r.t task number. Concretely, VMT-Adapter shares theknowledge from multiple tasks to enhance cross-task interaction while preservestask-specific knowledge via independent knowledge extraction modules. Notably,since task-specific modules require few parameters, VMT-Adapter can handle anarbitrary number of tasks with a negligible increase of trainable parameters.We also propose VMT-Adapter-Lite, which further reduces the trainableparameters by learning shared parameters between down- and up-projections.Extensive experiments on four dense scene understanding tasks demonstrate thesuperiority of VMT-Adapter(-Lite), achieving a 3.96%(1.34%) relativeimprovement compared to single-task full fine-tuning, while utilizing merely~1% (0.36%) trainable parameters of the pre-trained model.

VMT-Adapter: Parameter-Efficient Transfer Learning for Multi-Task Dense Scene Understanding