Datasets:
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-nc-4.0
|
| 3 |
+
task_categories:
|
| 4 |
+
- text-to-audio
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
tags:
|
| 8 |
+
- music
|
| 9 |
+
- chinese-traditional
|
| 10 |
+
- benchmark
|
| 11 |
+
size_categories:
|
| 12 |
+
- 100M<n<1G
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# CTIM-Bench: A Benchmark for Traditional Chinese Instrument Music Generation
|
| 16 |
+
|
| 17 |
+
This dataset serves as the official evaluation benchmark for the **CTIM-Gen** project (Submitted to ICMR 2026).
|
| 18 |
+
|
| 19 |
+
## Overview
|
| 20 |
+
|
| 21 |
+
**CTIM-Bench** contains **500 high-quality audio samples** selected to evaluate the generation quality of traditional Chinese instruments. It is used to compute metrics such as FAD, KL Divergence, and CLAP Score.
|
| 22 |
+
|
| 23 |
+
## Structure
|
| 24 |
+
|
| 25 |
+
- **`audio/`**: Contains 500 `.wav` files (e.g., `yangqin0007_s019.wav`).
|
| 26 |
+
- **`metadata.json`**: Contains the text prompts and instrument labels for each sample.
|
| 27 |
+
|
| 28 |
+
## Usage
|
| 29 |
+
|
| 30 |
+
This benchmark is designed to be used with the evaluation scripts provided in the [CTIM-Gen Model Repository](https://huggingface.co/CTIM-Gen/CTIM-Gen).
|
| 31 |
+
|
| 32 |
+
```python
|
| 33 |
+
# Example prompt from metadata.json
|
| 34 |
+
"A performance of yangqin, featuring bright and metallic tone, traditional Chinese folk music style, fast tempo."
|
| 35 |
+
```
|