Music genre classification is a field of study within the broader domain of Music Information Retrieval (MIR) that is still an open problem. This study aims at classifying music by Nigerian artists into respective genres using Convolutional Neural Networks (CNNs) and audio features extracted from the songs. To achieve this, a dataset of 524 Nigerian songs was collected from different genres. Each downloaded music file was converted from standard MP3 to WAV format and then trimmed to 30 seconds. The Librosa sc library was used for the analysis, visualization and further pre-processing of the music file which includes converting the audio signals to Mel-frequency cepstral coefficients (MFCCs). The MFCCs were obtained by taking performing a Discrete Cosine Transform on the logarithm of the Mel-scale filtered power spectrum of the audio signals. CNN architecture with multiple convolutional and pooling layers was used to learn the relevant features and classify the genres. Six models were trained using a categorical cross-entropy loss function with different learning rates and optimizers. Performance of the models was evaluated using accuracy, precision, recall, and F1-score. The models returned varying results from the classification experiments but model 3 which was trained with an Adagrad optimizer and learning rate of 0.01 had accuracy and recall of 75.1% and 84%, respectively. The results from the study demonstrated the effectiveness of MFCC and CNNs in music genre classification particularly with indigenous Nigerian artists.