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

First person action recognition is an increasingly researched topic becauseof the growing popularity of wearable cameras. This is bringing to lightcross-domain issues that are yet to be addressed in this context. Indeed, theinformation extracted from learned representations suffers from an intrinsicenvironmental bias. This strongly affects the ability to generalize to unseenscenarios, limiting the application of current methods in real settings wheretrimmed labeled data are not available during training. In this work, wepropose to leverage over the intrinsic complementary nature of audio-visualsignals to learn a representation that works well on data seen during training,while being able to generalize across different domains. To this end, weintroduce an audio-visual loss that aligns the contributions from the twomodalities by acting on the magnitude of their feature norm representations.This new loss, plugged into a minimal multi-modal action recognitionarchitecture, leads to strong results in cross-domain first person actionrecognition, as demonstrated by extensive experiments on the popularEPIC-Kitchens dataset.

Cross-Domain First Person Audio-Visual Action Recognition through Relative Norm Alignment