<![CDATA[This research examines the development of music genres and technological applications in music genre recognition through the MIR (Music Information Retrieval) approach. Automatic music genre labeling is expected to help, reduce, and suppress the role of humans in terms of music genre labeling. This research proposes the use of Mel Spectrogram as an audio representation in the frequency domain as well as Convolutional Neural Network (CNN), specifically the Inception V3 architecture. CNN was chosen for its ability to recognize complex and hierarchical patterns, which corresponds to the musical features represented in the spectrogram. Transfer learning techniques and fine-tuning of models trained on large datasets were applied, which allowed to improve accuracy. This study uses a dataset of 1000 audio files in .wav format, with each genre represented by 100 files, to evaluate the performance and effectiveness of the proposed method in the context of music genre classification.]]>
Copyrights © 2025
