Task-aware multiple instance learning for stress detection from facial video data

doi:10.1016/j.jad.2026.121472

Task-aware multiple instance learning for stress detection from facial video data

2026-02-20

0

PRE

AI

DOI：10.1016/j.jad.2026.121472

Original

Original request for help

Save

Abstract

En 中文

• Video-based facial cues enable non-invasive detection of acute stress. • Learning from weak labels identifies brief stress moments in long videos. • Combining top- and bottom-k video segments improves stress classification. • Attention modelling captures short- and long-term facial stress patterns. • Conditioning on speech activity improves stress detection accuracy.

Keywords:

stress detection

facial video analysis

multiple instance learning

weak labels

attention modelling

Task-aware multiple instance learning for stress detection from facial video data

Abstract

Journal

Organization