BROWSE
Detail
Conference Paper
SAGE: Stage-Adaptive Guided Reliability Modeling for Continuous Valence-Arousal Estimation
Cited - time in 
Share 
Share
- Authors
- Yubeen Lee, Sangeun Lee, Junyeop Cha, Eunil Park
- Issue Date
- 2026-06
- Citation
- Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 2026, pp.5312-5319
- Publisher
- Computer Vision Foundation
- Language
- English
- Type
- Conference Paper
- Abstract
- Continuous valence-arousal estimation in real-world envi- ronments is challenging due to inconsistent modality relia- bility and interaction-dependent variability in audio-visual signals. Existing approaches primarily focus on model- ing temporal dynamics, often overlooking the fact that modality reliability can vary substantially across interac- tion stages. To address this issue, we propose SAGE, a Stage-Adaptive reliability modeling framework that explic- itly estimates and calibrates modality-wise confidence dur- ing multimodal integration. SAGE introduces a reliability- aware fusion mechanism that dynamically rebalances au- dio and visual representations according to their stage- dependent informativeness, preventing unreliable signals from dominating the prediction process. By separating reli- ability estimation from feature representation, the proposed framework enables more stable emotion estimation under cross-modal noise, occlusion, and varying interaction con- ditions. Extensive experiments on the Aff-Wild2 benchmark demonstrate that SAGE consistently improves concordance correlation coefficient scores compared with existing multi- modal fusion approaches, highlighting the effectiveness of reliability-driven modeling for continuous affect prediction.
- KSP Keywords
- Ability Estimation, Affect prediction, Audio-visual, Correlation Coefficient, Emotion estimation, Existing Approaches, Feature Representation, Fusion mechanism, Multimodal integration, Prediction process, Real-world