Erhu Performance Technique Recognition Model

Introduction

The Erhu Performance Technique Recognition Model is an advanced tool designed to analyze erhu music, a traditional Chinese instrument, using deep learning. The model distinguishes 11 basic playing techniques such as split bow, pad bow, overtone, and others by analyzing acoustic characteristics. This model aids music information retrieval, education, and research, enhancing the study of erhu performance and contributing to the preservation and innovation of traditional music.

Architecture

The model utilizes time-frequency conversion, feature extraction, and pattern recognition to accurately categorize complex erhu techniques. It employs deep learning methodologies to achieve high accuracy in recognizing these performance techniques.

Training

Training involves fine-tuning a Swin-T model using mel spectrograms, with results indicating effective learning as shown by the loss curve, accuracy graphs, and a confusion matrix.

Guide: Running Locally

1. Download the Model:

   from modelscope import snapshot_download
   model_dir = snapshot_download('ccmusic-database/erhu_playing_tech')
   

2. Clone the Repository:

   git clone git@hf.co:ccmusic-database/erhu_playing_tech
   cd erhu_playing_tech
   

3. Setup Environment:
   Ensure all dependencies are installed and the environment is configured for running the model.

4. Cloud GPUs:
   For enhanced performance, consider using cloud GPU services like AWS, Google Cloud, or Azure.

License

The model is licensed under the MIT License, allowing for free use, modification, and distribution.