Latent-PER: ICA-Latent Code Editing Framework for Portrait Emotion Recognition

Although real-image emotion recognition has been developed in several studies, an acceptable accuracy level not achieved portrait drawings. This paper proposes a framework based on independent component analysis (ICA) and latent codes to overcome the performance degradation problem employs code extracted through generative adversarial network (GAN)-based encoder. It learns independently from factors that interfere with expression recognition, such as color, small occlusion, various face angles. is robust against environmental since it filters by adding emotion-relevant extractor extract only information related facial expressions code. In addition, image generated changing direction of eigenvector for each obtained ICA method. Since position changed, there little external change changes desired direction. technique helpful qualitative quantitative emotional learning. The experimental results reveal proposed model performs better than existing models, editing used this process suggests novel manipulation method ICA. Moreover, can be applied applications manipulation, automation subtitle production visually impaired, understanding emotions objects famous classic artwork, animation assistance.