RA-FER: Retrieval-Augmented Framework for Facial Expression Recognition

Haoran Xu; Yu Gao; Weihong Ren; Zhiyong Wang; Honghai Liu

doi:10.1016/j.ifacol.2025.12.456

RA-FER: Retrieval-Augmented Framework for Facial Expression Recognition

Haoran Xu
, Yu Gao
, Weihong Ren
, Zhiyong Wang
, Honghai Liu

Harbin Institute of Technology Shenzhen

Research output: Contribution to journal › Conference article › peer-review

Abstract

Facial Expression Recognition (FER) is challenging due to the subtle and ambiguous nature of human expressions. To address this issue, we propose RA-FER, a retrieval-augmented framework that enhances representation learning by leveraging semantically similar images from a retrieval gallery Specifically, RA-FER employs a dual cross-attention mechanism to integrate complementary information from both the query and the retrieved samples, while a router adaptively selects the most informative features for expression prediction. Extensive experiments on three real-world benchmarks demonstrate the robustness and superior performance of RA-FER.

Original language	English
Pages (from-to)	85-90
Number of pages	6
Journal	IFAC-PapersOnLine
Volume	59
Issue number	35
DOIs	https://doi.org/10.1016/j.ifacol.2025.12.456
State	Published - 2025
Externally published	Yes
Event	16th IFAC Symposium on Analysis, Design and Evaluation of Human-Machine Systems, HMS 2025 - Beijing, China Duration: 18 Nov 2025 → 21 Nov 2025

Keywords

Computer Vision
Cross-Attention
Facial Expression Recognition
Feature Enhancement
Retrieval-Augmented Model

Access to Document

10.1016/j.ifacol.2025.12.456

Cite this

@article{6c688d21db9c40118141eb4620239f32,

title = "RA-FER: Retrieval-Augmented Framework for Facial Expression Recognition",

abstract = "Facial Expression Recognition (FER) is challenging due to the subtle and ambiguous nature of human expressions. To address this issue, we propose RA-FER, a retrieval-augmented framework that enhances representation learning by leveraging semantically similar images from a retrieval gallery Specifically, RA-FER employs a dual cross-attention mechanism to integrate complementary information from both the query and the retrieved samples, while a router adaptively selects the most informative features for expression prediction. Extensive experiments on three real-world benchmarks demonstrate the robustness and superior performance of RA-FER.",

keywords = "Computer Vision, Cross-Attention, Facial Expression Recognition, Feature Enhancement, Retrieval-Augmented Model",

author = "Haoran Xu and Yu Gao and Weihong Ren and Zhiyong Wang and Honghai Liu",

note = "Publisher Copyright: {\textcopyright} 2025 Elsevier B.V.. All rights reserved.; 16th IFAC Symposium on Analysis, Design and Evaluation of Human-Machine Systems, HMS 2025 ; Conference date: 18-11-2025 Through 21-11-2025",

year = "2025",

doi = "10.1016/j.ifacol.2025.12.456",

language = "英语",

volume = "59",

pages = "85--90",

journal = "IFAC-PapersOnLine",

issn = "2405-8971",

publisher = "Elsevier B.V.",

number = "35",

}

TY - JOUR

T1 - RA-FER

T2 - 16th IFAC Symposium on Analysis, Design and Evaluation of Human-Machine Systems, HMS 2025

AU - Xu, Haoran

AU - Gao, Yu

AU - Ren, Weihong

AU - Wang, Zhiyong

AU - Liu, Honghai

PY - 2025

Y1 - 2025

N2 - Facial Expression Recognition (FER) is challenging due to the subtle and ambiguous nature of human expressions. To address this issue, we propose RA-FER, a retrieval-augmented framework that enhances representation learning by leveraging semantically similar images from a retrieval gallery Specifically, RA-FER employs a dual cross-attention mechanism to integrate complementary information from both the query and the retrieved samples, while a router adaptively selects the most informative features for expression prediction. Extensive experiments on three real-world benchmarks demonstrate the robustness and superior performance of RA-FER.

AB - Facial Expression Recognition (FER) is challenging due to the subtle and ambiguous nature of human expressions. To address this issue, we propose RA-FER, a retrieval-augmented framework that enhances representation learning by leveraging semantically similar images from a retrieval gallery Specifically, RA-FER employs a dual cross-attention mechanism to integrate complementary information from both the query and the retrieved samples, while a router adaptively selects the most informative features for expression prediction. Extensive experiments on three real-world benchmarks demonstrate the robustness and superior performance of RA-FER.

KW - Computer Vision

KW - Cross-Attention

KW - Facial Expression Recognition

KW - Feature Enhancement

KW - Retrieval-Augmented Model

UR - https://www.scopus.com/pages/publications/105029593903

U2 - 10.1016/j.ifacol.2025.12.456

DO - 10.1016/j.ifacol.2025.12.456

M3 - 会议文章

AN - SCOPUS:105029593903

SN - 2405-8971

VL - 59

SP - 85

EP - 90

JO - IFAC-PapersOnLine

JF - IFAC-PapersOnLine

IS - 35

Y2 - 18 November 2025 through 21 November 2025

ER -

RA-FER: Retrieval-Augmented Framework for Facial Expression Recognition

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this