CultureMERT: Continual Pre-Training for Cross-Cultural Music Representation Learning
Paper
•
2506.17818
•
Published
•
3
CultureMERT
A collection of model checkpoints from the paper "CultureMERT: Continual Pre-Training for Cross-Cultural Music Representation Learning".
Note Accepted at ISMIR 2025.
ntua-slp/CultureMERT-95M
Audio Classification • Updated • 28Note The multi-culturally adapted MERT model, continually pre-trained on 650 hours of Greek, Turkish, and Indian music traditions.
ntua-slp/CultureMERT-TA-95M
Audio Classification • Updated • 13Note The multi-cultural merged MERT model, created via task arithmetic by combining single-culture adaptations, each continually pre-trained on a distinct music tradition (Greek traditional/folk, Turkish Makam, Hindustani, and Carnatic).