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

Scene understanding using multi-modal data is necessary in many applications,e.g., autonomous navigation. To achieve this in a variety of situations,existing models must be able to adapt to shifting data distributions withoutarduous data annotation. Current approaches assume that the source data isavailable during adaptation and that the source consists of paired multi-modaldata. Both these assumptions may be problematic for many applications. Sourcedata may not be available due to privacy, security, or economic concerns.Assuming the existence of paired multi-modal data for training also entailssignificant data collection costs and fails to take advantage of widelyavailable freely distributed pre-trained uni-modal models. In this work, werelax both of these assumptions by addressing the problem of adapting a set ofmodels trained independently on uni-modal data to a target domain consisting ofunlabeled multi-modal data, without having access to the original sourcedataset. Our proposed approach solves this problem through a switchingframework which automatically chooses between two complementary methods ofcross-modal pseudo-label fusion -- agreement filtering and entropy weighting --based on the estimated domain gap. We demonstrate our work on the semanticsegmentation problem. Experiments across seven challenging adaptation scenariosverify the efficacy of our approach, achieving results comparable to, and insome cases outperforming, methods which assume access to source data. Ourmethod achieves an improvement in mIoU of up to 12% over competing baselines.Our code is publicly available at https://github.com/csimo005/SUMMIT.

SUMMIT: Source-Free Adaptation of Uni-Modal Models to Multi-Modal Targets