Deep learning algorithms have advanced music genre classification, enhancing user experiences and content organization in digital music platforms. This study assesses three deep learning models’ performance on the GTZAN dataset, a well-known standard for classifying musical genres: CNN-MLP, CNN-LSTM, and ResNet50. To improve model input quality, we employ feature engineering techniques, specifically extracting Mel-spectrograms and chroma features. The audio data is preprocessed through normalization and augmented using time-stretching and pitch-shifting to increase training sample diversity. The models undergo training and validation, with hyperparameter tuning conducted through grid search to optimize performance. Classification accuracy, precision, recall, and F1-score are used to evaluate the model’s effectiveness. Results show that ResNet achieves 90.2% accuracy and 89.8% precision, while CNN-LSTM records 88.3% accuracy and 87.7% precision. The CNN-MLP model performs slightly lower, with 86.0% accuracy and 85.4% precision. These findings indicate that all models are effective, with ResNet demonstrating superior performance. Future research should explore advanced network architectures and utilize diverse datasets to further enhance classification, particularly for complex and varied music genres.