Emolip: An Affective Computing Way Based on Image-Language Pre-Trained Model
Dayu Wu, Junjian Huang, Zhenquan Tang · 2024
This paper presents Emolip, an effective emotion recognition way based on Image-Language Pre-trained Model, designed to reduce costs and computational complexity. Emolip leverages a Image-Language Pre-training approach. The model utilizes a linear attention mechanism to enhance efficiency, outperforming traditional methods by reducing computational time. The experimental results, based on the EMOTIC dataset, demonstrate a significant improvement in recognition speed.