Facial expression recognition (FER) is significant in many application scenarios, such as driving scenarios with very different lighting conditions between day and night. Existing methods primarily focus on eliminating the negative effects of pose and identity information on FER, but overlook the challenges posed by lighting variations. So, this work proposes an efficient illuminationâinvariant dynamic FER method. To augment the robustness of FER methods to illumination variance, contrast normalisation is introduced to form a lowâlevel illuminationâinvariant expression features learningmodule. In addition, to extract dynamic and salient expression features, a twoâstage temporal attention mechanism is introduced to form a highâlevel dynamic salient expression features learning module deciphering dynamic facial expression patterns. Furthermore, additive angular margin loss is incorporated into the training of the proposed model to increase the distances between samples of different categories while reducing the distances between samples belonging to the same category. We conducted comprehensive experiments using the OuluâCASIA and DFEW datasets. On the OuluâCASIA VIS and NIR subsets in the normal illumination, the proposed method achieved accuracies of 92.08% and 91.46%, which are 1.01 and 7.06 percentage points higher than the SOTA results from the DCBLSTM and CELDL method, respectively. Based on the OuluâCASIA NIR subset in the dark illumination, the proposed method achieved an accuracies of 91.25%, which are 4.54 percentage points higher than the SOTA result from the CDLLNet method. On the DFEW dataset, the proposed method achieved a UAR of 60.67% and a WAR of 71.48% with 12M parameters, approaching the SOTA result from the VideoMAE model with 86M parameters. The outcomes of our experiments validate the effectiveness of the proposed dynamic FER method, affirming its ability in addressing the challenges posed by diverse illumination conditions in driving scenarios.
Pei, E, Guo, M, Berenguer, AD, He, L & Chen, H 2025, 'An Efficient IlluminationâInvariant Dynamic Facial Expression Recognition for Driving Scenarios', IET Intelligent Transport Systems, vol. 19, no. 1, e70009, pp. 1-16. https://doi.org/10.1049/itr2.70009
Pei, E., Guo, M., Berenguer, A. D., He, L., & Chen, H. (2025). An Efficient IlluminationâInvariant Dynamic Facial Expression Recognition for Driving Scenarios. IET Intelligent Transport Systems, 19(1), 1-16. Article e70009. https://doi.org/10.1049/itr2.70009
@article{aca2ab0bd27f4ec1a86b0a34cb8b83b0,
title = "An Efficient IlluminationâInvariant Dynamic Facial Expression Recognition for Driving Scenarios",
abstract = "Facial expression recognition (FER) is significant in many application scenarios, such as driving scenarios with very different lighting conditions between day and night. Existing methods primarily focus on eliminating the negative effects of pose and identity information on FER, but overlook the challenges posed by lighting variations. So, this work proposes an efficient illuminationâinvariant dynamic FER method. To augment the robustness of FER methods to illumination variance, contrast normalisation is introduced to form a lowâlevel illuminationâinvariant expression features learningmodule. In addition, to extract dynamic and salient expression features, a twoâstage temporal attention mechanism is introduced to form a highâlevel dynamic salient expression features learning module deciphering dynamic facial expression patterns. Furthermore, additive angular margin loss is incorporated into the training of the proposed model to increase the distances between samples of different categories while reducing the distances between samples belonging to the same category. We conducted comprehensive experiments using the OuluâCASIA and DFEW datasets. On the OuluâCASIA VIS and NIR subsets in the normal illumination, the proposed method achieved accuracies of 92.08% and 91.46%, which are 1.01 and 7.06 percentage points higher than the SOTA results from the DCBLSTM and CELDL method, respectively. Based on the OuluâCASIA NIR subset in the dark illumination, the proposed method achieved an accuracies of 91.25%, which are 4.54 percentage points higher than the SOTA result from the CDLLNet method. On the DFEW dataset, the proposed method achieved a UAR of 60.67% and a WAR of 71.48% with 12M parameters, approaching the SOTA result from the VideoMAE model with 86M parameters. The outcomes of our experiments validate the effectiveness of the proposed dynamic FER method, affirming its ability in addressing the challenges posed by diverse illumination conditions in driving scenarios.",
author = "Ercheng Pei and Man Guo and Berenguer, {Abel D{\'i}az} and Lang He and HaiFeng Chen",
note = "Publisher Copyright: {\textcopyright} 2025 The Author(s). IET Intelligent Transport Systems published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.",
year = "2025",
month = mar,
day = "4",
doi = "10.1049/itr2.70009",
language = "English",
volume = "19",
pages = "1--16",
journal = "IET Intelligent Transport Systems",
issn = "1751-956X",
publisher = "Institution of Engineering and Technology",
number = "1",
}