Efficient Feature Extraction and Late Fusion Strategy for Audiovisual Emotional Mimicry Intensity Estimation

arXiv:2403.11757v1 Announce Type: cross
Abstract: In this paper, we present the solution to the Emotional Mimicry Intensity (EMI) Estimation challenge, which is part of 6th Affective Behavior Analysis in-the-wild (ABAW) Competition.The EMI Estimation challenge task aims to evaluate the emotional intensity of seed videos by assessing them from a set of predefined emotion categories (i.e., "Admiration," "Amusement," "Determination," "Empathic Pain," "Excitement," and "Joy").

abstract analysis arxiv behavior behavior analysis challenge competition cs.lg cs.mm cs.sd eess.as extraction feature feature extraction fusion intensity mimicry paper part seed solution strategy them type videos