UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

When Facial Expression Recognition Meets Few-Shot Learning: A Joint and Alternate Learning Framework

Zou, X; Yan, Y; Xue, J; Chen, S; Wang, H; (2022) When Facial Expression Recognition Meets Few-Shot Learning: A Joint and Alternate Learning Framework. In: Proceedings of the Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI-22). AAAI (In press). Green open access

[thumbnail of 2201.06781.pdf]
Preview
Text
2201.06781.pdf - Accepted Version

Download (1MB) | Preview

Abstract

Human emotions involve basic and compound facial expres- sions. However, current research on facial expression recog- nition (FER) mainly focuses on basic expressions, and thus fails to address the diversity of human emotions in practical scenarios. Meanwhile, existing work on compound FER re- lies heavily on abundant labeled compound expression train- ing data, which are often laboriously collected under the pro- fessional instruction of psychology. In this paper, we study compound FER in the cross-domain few-shot learning set- ting, where only a few images of novel classes from the target domain are required as a reference. In particular, we aim to identify unseen compound expressions with the model trained on easily accessible basic expression datasets. To alleviate the problem of limited base classes in our FER task, we propose a novel Emotion Guided Similarity Network (EGS-Net), con- sisting of an emotion branch and a similarity branch, based on a two-stage learning framework. Specifically, in the first stage, the similarity branch is jointly trained with the emo- tion branch in a multi-task fashion. With the regularization of the emotion branch, we prevent the similarity branch from overfitting to sampled base classes that are highly overlapped across different episodes. In the second stage, the emotion branch and the similarity branch play a “two-student game” to alternately learn from each other, thereby further improving the inference ability of the similarity branch on unseen com- pound expressions. Experimental results on both in-the-lab and in-the-wild compound expression datasets demonstrate the superiority of our proposed method against several state- of-the-art methods.

Type: Proceedings paper
Title: When Facial Expression Recognition Meets Few-Shot Learning: A Joint and Alternate Learning Framework
Event: Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI-22)
Dates: 22 February 2022 - 01 March 2022
Open access status: An open access version is available from UCL Discovery
Publisher version: https://aaai.org/Conferences/AAAI-22/
Language: English
Additional information: This version is the author accepted manuscript. For information on re-use, please refer to the publisher's terms and conditions.
UCL classification: UCL
UCL > Provost and Vice Provost Offices > UCL BEAMS
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Maths and Physical Sciences
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Maths and Physical Sciences > Dept of Statistical Science
URI: https://discovery.ucl.ac.uk/id/eprint/10142364
Downloads since deposit
33Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item