Music plays a major role in daily life and serves as a key means for expressing human emotions. Automatic classification of Iranian traditional music tracks is a fascinating yet challenging subject, particularly for those interested in Iranian music dastgahs (subgenres).This paper proposes a novel method for Iranian traditional music genre recognition using Persian music tracks. Six Iranian music genres, namely Shour, Nava, Mahour, Segah, Chahargah, and Homayoun, are considered. To accurately detect genres, convolutional neural networks (CNNs), one-dimensional convolutional neural networks (1DCNNs), and long short-term memories (LSTMs) are employed.All models are fed extracted pitch features, with the music pitch converted into a sequential note vector and a visual representation in the form of a graph illustrating the musical structure. Finally, an ensemble model combines the predictions from all models.The proposed approach is evaluated using the "Arg" database, which includes solo melodic instrument tracks with no limitations on playing style, instrument, tempo, or techniques. The proposed method achieved a recognition accuracy of 77.35%, which improved to 80.44% with the use of data augmentation techniques.The experimental results, including accuracy, F1 score, and standard deviation (STD), demonstrate the effectiveness of the approach, showing better performance compared to other methods for Iranian genre recognition.