Emotion-Aware Music Recommendations: Evaluating Custom CNN vs. VGG16 and InceptionV3
Varesh Patel, Khyati Mehta · 2024
This study conducts a comparative analysis of various pretrained models for detecting emotions from facial expressions, introducing a novel Convolutional Neural Network (CNN) designed to enhance accuracy. We evaluate the performance of well-known models such as VGG16 and InceptionV3 alongside our proposed CNN architecture to determine the most effective model for incorporating facial emotion detection into music recommendation systems. Our system utilizes facial expressions captured through webcams or images to deliver personalized music suggestions that reflect users' current emotional states, aiming to enhance user engagement on music streaming platforms. By outlining methodology, including dataset compilation and preprocessing strategies, we test and validate our CNN architecture, revealing its superiority over existing models in accuracy. The integration of sophisticated facial emotion detection into recommendation systems offers a transformative way to improve the user experience by providing emotionally attuned music suggestions. Our findings demonstrate that this advancement holds promise for applications in digital music streaming, enhancing user engagement and personalization. By leveraging facial expression analysis, we can create more effective music recommendation systems that cater to individual users' emotional needs, leading to increased user satisfaction and loyalty.