Deep Learning-Based Detection and Classification of Traditional South Asian Musical Instruments Using Convolutional Neural Networks (CNNs)

This research explores the application of deep learning techniques to the classification and identification of classical South Asian musical instruments. South Asia has a rich and diverse musical heritage, comprising instruments with distinctive acoustic characteristics, structures, and playing techniques. Automatically identifying these instruments presents an interesting challenge in music information retrieval and audio-related classification research.

The project investigates a convolutional neural network (CNN)-based approach implemented using TensorFlow and its Keras API to classify musical instrument samples across 15 categories. The model architecture consists of four convolutional layers with 32, 64, 128, and 256 filters, respectively, incorporating ReLU activation and max-pooling operations. A sigmoid output layer is used to estimate class probabilities, enabling multi-label classification.

The model was trained using a dataset of 6,000 samples, split into 80% training and 20% testing data. Training was conducted using Google Colab with an NVIDIA T4 GPU for 10 epochs, with a batch size of 32, a learning rate of 0.001, and binary cross-entropy as the loss function.

The work establishes a baseline deep learning approach for the automated classification of classical South Asian musical instruments and provides a foundation for further exploration of music recognition systems, including potential real-time instrument identification during musical performances.