Emotion-Aware Music Recommendation

It is common to listen songs that match one's mood. Thus, an AI music recommendation system aware of the user's emotions likely provide a superior user experience one unaware. In this paper, we present emotion-aware system. Multiple models are discussed and evaluated for affect identification from live image user. We propose two models: DRViT, which applies dynamic routing vision transformers, InvNet50, uses involution. All considered trained on AffectNet dataset. Each model outputs estimated valence arousal under circumplex affect. These values compared in Spotify dataset, top-five closest-matching presented Experimental results testing presented.