Cumulative attributes for pain intensity estimation

Pain estimation from face video is a hard problem in automatic behaviour understanding. One major obstacle is the di culty of collecting sufficient amounts of data, with balanced amounts of data for all pain intensity levels. To overcome this, we propose to adopt Cumulative Attributes, which assume...

Full description

Bibliographic Details
Main Authors: Joy, Egede, Michel, Valstar
Format: Conference or Workshop Item
Published: 2017
Subjects:
Online Access:https://eprints.nottingham.ac.uk/45830/
_version_ 1848797201556307968
author Joy, Egede
Michel, Valstar
author_facet Joy, Egede
Michel, Valstar
author_sort Joy, Egede
building Nottingham Research Data Repository
collection Online Access
description Pain estimation from face video is a hard problem in automatic behaviour understanding. One major obstacle is the di culty of collecting sufficient amounts of data, with balanced amounts of data for all pain intensity levels. To overcome this, we propose to adopt Cumulative Attributes, which assume that attributes for high pain levels with few examples are a superset of all attributes of lower pain levels. Experimental results show a consistent relative performance increase in the order of 20% regardless of features used. Our final system significantly outperforms the state of the art on the UNBC McMaster Shoulder Pain database by using cumulative attributes with Relevance Vector Regression on a combination of features, including appearance, geometric, and deep learned features.
first_indexed 2025-11-14T20:00:07Z
format Conference or Workshop Item
id nottingham-45830
institution University of Nottingham Malaysia Campus
institution_category Local University
last_indexed 2025-11-14T20:00:07Z
publishDate 2017
recordtype eprints
repository_type Digital Repository
spelling nottingham-458302020-05-04T19:17:29Z https://eprints.nottingham.ac.uk/45830/ Cumulative attributes for pain intensity estimation Joy, Egede Michel, Valstar Pain estimation from face video is a hard problem in automatic behaviour understanding. One major obstacle is the di culty of collecting sufficient amounts of data, with balanced amounts of data for all pain intensity levels. To overcome this, we propose to adopt Cumulative Attributes, which assume that attributes for high pain levels with few examples are a superset of all attributes of lower pain levels. Experimental results show a consistent relative performance increase in the order of 20% regardless of features used. Our final system significantly outperforms the state of the art on the UNBC McMaster Shoulder Pain database by using cumulative attributes with Relevance Vector Regression on a combination of features, including appearance, geometric, and deep learned features. 2017-11-13 Conference or Workshop Item PeerReviewed Joy, Egede and Michel, Valstar (2017) Cumulative attributes for pain intensity estimation. In: 19th ACM International Conference on Multimodal Interaction, 13-17 November 2017, Glasgow, Scotland. (In Press) Pain estimation; Attribute learning; Multi-output regression; Relevance Vector Machines (RVM)
spellingShingle Pain estimation; Attribute learning; Multi-output regression; Relevance Vector Machines (RVM)
Joy, Egede
Michel, Valstar
Cumulative attributes for pain intensity estimation
title Cumulative attributes for pain intensity estimation
title_full Cumulative attributes for pain intensity estimation
title_fullStr Cumulative attributes for pain intensity estimation
title_full_unstemmed Cumulative attributes for pain intensity estimation
title_short Cumulative attributes for pain intensity estimation
title_sort cumulative attributes for pain intensity estimation
topic Pain estimation; Attribute learning; Multi-output regression; Relevance Vector Machines (RVM)
url https://eprints.nottingham.ac.uk/45830/