GuitarNeXt: An Advanced Convolutional Neural Network Architecture for Music Genre Classification

Music genre classification is a challenging task that has been extensively addressed using various deep learning methods.Recently, convolutional neural networks (CNNs) have shown significant promise in this domain.This paper introduces GuitarNeXt, a novel CNN architecture designed specifically for music genre classification.Our approach utilizes a publicly available dataset containing audio recordings and spectral images across multiple genres to evaluate the performance of GuitarNeXt.The architecture of GuitarNeXt includes four primary layers: a stem layer employing patchify convolution to produce initial tensors, a main GuitarNeXt layer that integrates a hybrid attention mechanism with depth concatenation and scaling convolutions, a downsampling layer combining average and maximum pooling with depth concatenation, and an output layer that applies global pooling to produce a feature map for classification.Experimental results demonstrate that GuitarNeXt achieves a classification accuracy of 96.40% and a precision of 96.59% on the test set, highlighting its effectiveness and potential as a robust tool for automated music genre classification.This innovative model not only advances the field of music analysis but also sets a new benchmark for subsequent research in the area.