Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning to Discover: A Generalized Framework for Raga Identification without Forgetting

Parampreet Singh, Somya Kumar, Chaitanya Shailendra Nitawe, Vipul Arora

TL;DR

The paper addresses open-set Raga identification in Indian Art Music by adapting Generalized Category Discovery to jointly learn from labeled and unlabeled audio, mitigating catastrophic forgetting while discovering unseen Ragа categories. A CNN–LSTM feature extractor feeds a unified embedding space optimized with both supervised and unsupervised contrastive losses, followed by clustering to form new Ragа groups. Across PIM and Saraga datasets, the proposed M2 method consistently outperforms a prior NCD baseline, achieving better accuracy and cluster quality while preserving knowledge of known Ragа classes. This approach advances IAM representation learning and offers scalable, cross-dataset Ragа discovery for large music archives; code will be released publicly.

Abstract

Raga identification in Indian Art Music (IAM) remains challenging due to the presence of numerous rarely performed Ragas that are not represented in available training datasets. Traditional classification models struggle in this setting, as they assume a closed set of known categories and therefore fail to recognise or meaningfully group previously unseen Ragas. Recent works have tried categorizing unseen Ragas, but they run into a problem of catastrophic forgetting, where the knowledge of previously seen Ragas is diminished. To address this problem, we adopt a unified learning framework that leverages both labeled and unlabeled audio, enabling the model to discover coherent categories corresponding to the unseen Ragas, while retaining the knowledge of previously known ones. We test our model on benchmark Raga Identification datasets and demonstrate its performance in categorizing previously seen, unseen, and all Raga classes. The proposed approach surpasses the previous NCD-based pipeline even in discovering the unseen Raga categories, offering new insights into representation learning for IAM tasks.

Learning to Discover: A Generalized Framework for Raga Identification without Forgetting

TL;DR

The paper addresses open-set Raga identification in Indian Art Music by adapting Generalized Category Discovery to jointly learn from labeled and unlabeled audio, mitigating catastrophic forgetting while discovering unseen Ragа categories. A CNN–LSTM feature extractor feeds a unified embedding space optimized with both supervised and unsupervised contrastive losses, followed by clustering to form new Ragа groups. Across PIM and Saraga datasets, the proposed M2 method consistently outperforms a prior NCD baseline, achieving better accuracy and cluster quality while preserving knowledge of known Ragа classes. This approach advances IAM representation learning and offers scalable, cross-dataset Ragа discovery for large music archives; code will be released publicly.

Abstract

Raga identification in Indian Art Music (IAM) remains challenging due to the presence of numerous rarely performed Ragas that are not represented in available training datasets. Traditional classification models struggle in this setting, as they assume a closed set of known categories and therefore fail to recognise or meaningfully group previously unseen Ragas. Recent works have tried categorizing unseen Ragas, but they run into a problem of catastrophic forgetting, where the knowledge of previously seen Ragas is diminished. To address this problem, we adopt a unified learning framework that leverages both labeled and unlabeled audio, enabling the model to discover coherent categories corresponding to the unseen Ragas, while retaining the knowledge of previously known ones. We test our model on benchmark Raga Identification datasets and demonstrate its performance in categorizing previously seen, unseen, and all Raga classes. The proposed approach surpasses the previous NCD-based pipeline even in discovering the unseen Raga categories, offering new insights into representation learning for IAM tasks.
Paper Structure (22 sections, 10 equations, 3 figures, 1 table, 1 algorithm)

This paper contains 22 sections, 10 equations, 3 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Dataset
  5. Feature Extractor
  6. Supervised Contrastive Loss
  7. Unsupervised Contrastive Loss
  8. Total Contrastive Objective
  9. Clustering Methods
  10. Evaluation Metrics
  11. Experiments
  12. Datasets
  13. Feature Extractor Selection
  14. Baseline (BL)
  15. Unsupervised Contrastive Learning (M1)
  16. ...and 7 more sections

Figures (3)

  • Figure 1: Overall pipeline of the proposed framework.
  • Figure 2: Architecture of the CNN-LSTM model $f_\theta$
  • Figure 3: Confusion Matrix for M2 on $D^u$ for PIM dataset