HamNava Dataset

HamNava (meaning "in unison" in Farsi) is a dataset designed for segment-level multi-label instrument classification in Iranian classical music. It consists of 6,000 five-second excerpts, each annotated for the presence of nine musical elements (eight instruments + vocals), using soft labels gathered via a large-scale, community-driven annotation effort on Telegram.

The dataset is available for non-commercial research purposes only. By requesting access, you agree to use the data exclusively for academic and research purposes. You can download a sample portion of the dataset from here. Please refer to the included README file from here for details on the dataset format, labels, and usage instructions. For the complete methodology and background, see the original paper: HamNava Paper. You might also refer to the GitHub repository for the implementation of the crowd-sourcing application.

How to Request Access:

You can request access via filing out the Google Form (Prefered), completing the below form, or email directly. Please fill in the necessary information and agree to the terms before submitting.

Full Name:

Institutional Email Address:

Purpose of Use:

Affiliated Institution / Organization:

Additional Notes:

Agreement: By submitting this request, you confirm that the dataset will be used solely for academic or research purposes. You agree not to use it for any commercial or non-research activities.

Submit Request

If above unavailable, you can request the dataset via email by sending afformnetioned details and agreeing to use it only for research purposes. With the subject HamNava Dataset Request send the email to mohseni0pouya\at\gmail\dot\com CC'ing babaali\at\ut\dot\ac\dot\ir.