IndRaga: An Open Hindustani Classical Music Dataset of Precomputed Audio Features

IndRaga is an open feature dataset derived from 526 Hindustani classical music recordings spanning 76 Ragas and 10 Thaats. It contains precomputed frame-level representations including pYIN and RMVPE fundamental frequency estimates, chroma features, and MFCCs, stored as aligned NumPy dictionaries at 16 kHz / 10 ms resolution. Metadata includes identifiers and URLs pointing to publicly available archival sources used for feature extraction. The release is intended for research in Raga recognition and melodic structure modeling.