Dataset Information

A performance comparison of eight commercially available automatic classifiers for facial affect recognition.

ABSTRACT: In the wake of rapid advances in automatic affect analysis, commercial automatic classifiers for facial affect recognition have attracted considerable attention in recent years. While several options now exist to analyze dynamic video data, less is known about the relative performance of these classifiers, in particular when facial expressions are spontaneous rather than posed. In the present work, we tested eight out-of-the-box automatic classifiers, and compared their emotion recognition performance to that of human observers. A total of 937 videos were sampled from two large databases that conveyed the basic six emotions (happiness, sadness, anger, fear, surprise, and disgust) either in posed (BU-4DFE) or spontaneous (UT-Dallas) form. Results revealed a recognition advantage for human observers over automatic classification. Among the eight classifiers, there was considerable variance in recognition accuracy ranging from 48% to 62%. Subsequent analyses per type of expression revealed that performance by the two best performing classifiers approximated those of human observers, suggesting high agreement for posed expressions. However, classification accuracy was consistently lower (although above chance level) for spontaneous affective behavior. The findings indicate potential shortcomings of existing out-of-the-box classifiers for measuring emotions, and highlight the need for more spontaneous facial databases that can act as a benchmark in the training and testing of automatic emotion recognition systems. We further discuss some limitations of analyzing facial expressions that have been recorded in controlled environments.

SUBMITTER: Dupre D

PROVIDER: S-EPMC7182192 | biostudies-literature | 2020

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A performance comparison of eight commercially available automatic classifiers for facial affect recognition.

Dupré Damien D Krumhuber Eva G EG Küster Dennis D McKeown Gary J GJ

PloS one 20200424 4

In the wake of rapid advances in automatic affect analysis, commercial automatic classifiers for facial affect recognition have attracted considerable attention in recent years. While several options now exist to analyze dynamic video data, less is known about the relative performance of these classifiers, in particular when facial expressions are spontaneous rather than posed. In the present work, we tested eight out-of-the-box automatic classifiers, and compared their emotion recognition perfo ...[more]

PMID: 32330178

Similar Datasets

Project description:ImportanceBotulinum toxin neuromodulators are an important treatment for facial synkinesis. Whether a difference in efficacy exists among the 3 different botulinum neuromodulators used in treating this condition remains unknown.ObjectiveTo evaluate the effectiveness of 3 commercially available botulinum toxin neuromodulators in the treatment of facial synkinesis.Design, setting, and participantsIn this single-blind, 3-arm comparison randomized clinical trial, 28 patients at the Facial Nerve Center, University of Utah, Salt Lake City, were randomized to onabotulinumtoxinA, abobotulinumtoxinA, or incobotulinumtoxinA treatment. Each patient was given the Synkinesis Assessment Questionnaire (SAQ) to assess severity of synkinesis before treatment and 1, 2, and 4 weeks after treatment, and improvements were compared among the groups. Data were collected from July 3, 2012, to March 31, 2015.InterventionsBotulinum toxin type A neuromodulator (onabotulinumtoxinA, abobotulinumtoxinA, or incobotulinumtoxinA) injected into synkinetic areas of the face.Main outcomes and measuresSynkinesis assessed using the SAQ (score range, 20-100; lower scores indicate less severe synkinesis) before treatment and 1, 2, and 4 weeks after treatment.ResultsA total of 28 patients (mean [SD] age, 49.1 [18.5] years; 8 [28.6%] male and 20 [71.4%] female), with 6 patients enrolled multiple times, received 38 treatments (15 onabotulinumtoxinA, 13 abobotulinumtoxinA, and 10 incobotulinumtoxinA). No significant difference existed in baseline pretreatment SAQ scores among the 3 groups. Mean (SD) SAQ score improvement at 4 weeks was 41% (31%) for the onabotulinumtoxinA, 42% (20%) for the abobotulinumtoxinA, and 17% (18%) for the incobotulinumtoxinA groups. No significant differences were noted in SAQ score improvements among the 3 groups at weeks 1 and 2 after treatment (week 1 mean improvements of 42% in the onabotulinumtoxinA, 45% in the abobotulinumtoxinA, and 26% in the incobotulinumtoxinA groups; P = .19; week 2 mean improvements of 43% in the onabotulinumtoxinA, 46% in the abobotulinumtoxinA, and 28% in the incobotulinumtoxinA groups; P = .20). The difference in mean SAQ score improvement for abobotulinumtoxinA vs incobotulinumtoxinA from pretreatment to 4 weeks after treatment was not significant (30 vs 12 points; P = .11) despite a significant difference in mean total SAQ score for abobotulinumtoxinA vs incobotulinumtoxinA (40.34 vs 58.00; P = .02).Conclusions and relevanceAbobotulinumtoxinA had similar efficacy to onabotulinumtoxinA and incobotulinumtoxinA for the management of facial synkinesis up to 4 weeks after treatment. IncobotulinumtoxinA had significantly less effect on SAQ score improvement than onabotulinumtoxinA at 4 weeks, perhaps because of the shorter duration of action. Shorter intervals between treatments or larger doses may be required when using incobotulinumtoxinA treatment for facial synkinesis.Trial registrationclinicaltrials.gov Identifier: NCT03048383.Level of evidence1.

Dataset Information

A performance comparison of eight commercially available automatic classifiers for facial affect recognition.

Publications

A performance comparison of eight commercially available automatic classifiers for facial affect recognition.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets