CTIS (Chinese Traditional Instrument Sound Model)
| Property | Value |
|---|---|
| Author | ccmusic-database |
| Best Performance | 98% (CQT), 97.3% (Mel) |
| Dataset Access | Available on HuggingFace |
| Paper | Upcoming in TISMIR 2025 |
What is CTIS?
CTIS is a groundbreaking AI model designed specifically for analyzing and classifying Chinese traditional musical instruments. Built on an extensive dataset encompassing over 200 different types of instruments, including rare and ethnic minority instruments, this model represents a significant advancement in musical instrument recognition technology.
Implementation Details
The model employs multiple backbone architectures, with RegNet_Y_32GF achieving the best performance (98% accuracy on CQT features). The implementation supports various audio feature extraction methods, including Mel-spectrograms, Constant-Q Transform (CQT), and Chroma features.
- Supports multiple neural architectures (VIT, Swin, RegNet, VGG19)
- Easy integration through ModelScope
- Comprehensive evaluation framework available
Core Capabilities
- Accurate identification of 200+ Chinese traditional instruments
- Support for modified and ethnic minority instruments
- High-performance across different audio feature types
- Robust classification with state-of-the-art accuracy
Frequently Asked Questions
Q: What makes this model unique?
CTIS is the first comprehensive model focusing specifically on Chinese traditional instruments, offering unprecedented coverage of rare and ethnic instruments with exceptional accuracy.
Q: What are the recommended use cases?
The model is ideal for music information retrieval, ethnomusicological research, digital preservation of cultural heritage, and educational applications in traditional Chinese music.
