Overview

This repository focuses on Meta-Learning-based Percussion Transcription and Tala Identification using low-resource audio data. The two primary tasks include:

1. Percussion Transcription:

   • Tabla Stroke Transcription (TST): Identifying and classifying Tabla strokes, along with their timing information,

     1. From Tabla solo recordings.
     2. From complete concert audio, including melodic instruments and the singer's voice.

   • Automatic Drum Transcription (ADT): Extracting drum strokes and their timing information from drum audio tracks.

     1. Drum Only Transcription (DTD): Transcription of drum strokes in isolation.
     2. Drums in Presence of Other Percussion (DTP): Transcribing drums in the presence of other percussion instruments.
     3. Drum in Melodic Instruments (DTM): Transcribing drums alongside melodic instruments.

2. Tala Identification:

   • Recognizing and analyzing rhythmic cycles (Tala) in Hindustani classical music.

---

To facilitate research in these areas, this repository includes:

1. Presentation: A PowerPoint file that explains key concepts of rhythm in Hindustani music, including the structure of Tala.

2. Synthetic_Mridangam_Stroke_Dataset_D_M: A curated dataset of synthetic Mridangam strokes for audio research and analysis.

---

Download "Synthetic_Mridangam_Stroke_Dataset_D_M"

---

The dataset contains synthetic audio samples of Mridangam strokes. The Mridangam, a prominent South Indian percussion instrument, is widely used in Carnatic music and has distinctive rhythmic patterns.

      Synthetic_Mridangam_Stroke_D_M/
  
          ├── Audio/            # Audio files for 10 classes of strokes
          
          ├── Onsets_10_Classes/     # Onset files for 10 stroke classes
          
          ├── Onsets_11_Classes/     # Onset files for 11 stroke classes (including silence class)
          
          └── README.md              # Additional details about the dataset

Each main folder contains subfolders structured as follows:
      o	  Audio, Onsets_10_Classes, Onsets_11_Classes:
      o	  Each subfolder consists of 6 tonic folders: B, C, C#, D, D#, and E.
      o	  Each tonic folder contains 120 files, totaling 720 files per main folder

---

*The codes will be uploaded soon.