A Comparative Study of DenseNets for Vietnamese Traditional Music Genre Classification

The rapid progress of artificial intelligence has led to a corresponding increase in the demand for classification of music genres. Numerous firms have enlisted architects with exceptional expertise for the execution of this project. However, it is crucial to recognize that Vietnamese traditional music, along with other types of Asian traditional music, has not attained the same degree of outstanding performance qualities as other musical genres. This study aimed to assess the effectiveness of two architectural designs, namely Late Fusion Convolutional Neural Networks and Densely Connected Convolutional Networks, through the utilization of diverse visual transformations of musical patterns. The ongoing inquiry involved doing a comparative study on a meticulously preserved dataset of Vietnamese traditional music. The results obtained from our research inquiry have shown the architectural design that is best suited for this particular undertaking. The results of our analysis indicate that Late Fusion Convolutional Neural Networks are a better suitable option for achieving this specific g oal. T his s tudy m akes a substantial contribution to the field of music information retrieval ( MIR) by investigating the effectiveness and precision of Densely Connected Convolutional Neural Networks (DenseNet)-based approaches in categorizing Vietnamese traditional music genres.