"Aff-Wild2: Extending the aff-wild database for affect recognition"

IEEE DataPort 2026 Dataset Open access

Read full text
Authors: Kollias Dimitrios
DOI: 10.21227/d5dw-yj11
Cited by: 0

Abstract

"The Aff-wild2 dataset consists of 558 YouTube videos, totaling 2,786,201 frames, each annotated with valence and arousal values (range [-1, +1]). The training, validation, and test subsets contain 351, 71, and 139 subjects, respectively."

Fields of Study & Subjects

Emotion and Mood Recognition Human Pose and Action Recognition Sentiment Analysis and Opinion Mining

Publication details

Journal / source
IEEE DataPort
Publication date
20 January 2026
Document type
Dataset