Add dataset card
Browse files
README.md
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
pretty_name: "ICASSP 2027 Audio Editing Challenge Track 1: Single Model Track"
|
| 3 |
+
task_categories:
|
| 4 |
+
- audio-to-audio
|
| 5 |
+
viewer: false
|
| 6 |
+
---
|
| 7 |
+
|
| 8 |
+
<p align="center">
|
| 9 |
+
<a href="https://audio-editing-challenge.github.io/">
|
| 10 |
+
<img src="https://raw.githubusercontent.com/Audio-Editing-Challenge/Audio-Editing-Challenge.github.io/main/img/audio-editing-challenge-logo-hd-4k.png" width="900" alt="ICASSP 2027 Audio Editing Challenge">
|
| 11 |
+
</a>
|
| 12 |
+
</p>
|
| 13 |
+
|
| 14 |
+
<h1 align="center">ICASSP 2027 Audio Editing Challenge Track 1: Single Model Track</h1>
|
| 15 |
+
|
| 16 |
+
<p align="center">
|
| 17 |
+
<a href="https://audio-editing-challenge.github.io/"><img src="https://img.shields.io/badge/Challenge-Website-2563EB?style=flat-square" alt="Challenge website"></a>
|
| 18 |
+
<a href="https://github.com/Audio-Editing-Challenge/Audio-Editing-Challenge-Baseline"><img src="https://img.shields.io/badge/Baseline-Code-181717?style=flat-square&logo=github" alt="Baseline code"></a>
|
| 19 |
+
<a href="https://huggingface.co/datasets/BoJack/MMAE"><img src="https://img.shields.io/badge/%F0%9F%A4%97-MMAE%20Dataset-FFD21E?style=flat-square" alt="MMAE dataset on Hugging Face"></a>
|
| 20 |
+
</p>
|
| 21 |
+
|
| 22 |
+
Test inputs for Track 1 of the [ICASSP 2027 Audio Editing Challenge](https://audio-editing-challenge.github.io/). The set has 500 examples. All of them are in the MMAE `single` complexity category and cover speech, music, sound, and their mixtures.
|
| 23 |
+
|
| 24 |
+
This release contains only input audio and instructions. The evaluation rubrics stay private until the official results are finalized.
|
| 25 |
+
|
| 26 |
+
## Files
|
| 27 |
+
|
| 28 |
+
```text
|
| 29 |
+
track1_single_model.json # metadata (id + messages)
|
| 30 |
+
wav/<id>/audio1.wav # input audio
|
| 31 |
+
```
|
| 32 |
+
|
| 33 |
+
```bash
|
| 34 |
+
hf download ICASSP2027-AudioEditingChallenge/Track1-Single-Model --repo-type dataset --local-dir track1
|
| 35 |
+
```
|
| 36 |
+
|
| 37 |
+
## Format
|
| 38 |
+
|
| 39 |
+
`messages` uses the same chat format as [MMAE](https://github.com/ddlBoJack/MMAE). Each `audio_url` is relative to the dataset root.
|
| 40 |
+
|
| 41 |
+
```json
|
| 42 |
+
{
|
| 43 |
+
"id": "<sample_id>",
|
| 44 |
+
"messages": [
|
| 45 |
+
{"role": "user", "content": [
|
| 46 |
+
{"type": "text", "text": "<editing instruction>"},
|
| 47 |
+
{"type": "audio", "audio_url": "wav/<sample_id>/audio1.wav"}
|
| 48 |
+
]}
|
| 49 |
+
]
|
| 50 |
+
}
|
| 51 |
+
```
|
| 52 |
+
|
| 53 |
+
## Submission
|
| 54 |
+
|
| 55 |
+
Each JSON entry is one example. Submit exactly one edited 16 kHz mono WAV per `id` (500 in total), plus a `submission.jsonl` with one line per example:
|
| 56 |
+
|
| 57 |
+
```json
|
| 58 |
+
{"id": "<sample_id>", "audio_path": "audio/<sample_id>.wav"}
|
| 59 |
+
```
|
| 60 |
+
|
| 61 |
+
Package `submission.jsonl` together with the `audio/` folder and upload it to the [leaderboard](https://audio-editing-challenge.github.io/leaderboard/). The full rules are on the [Track 1](https://audio-editing-challenge.github.io/track1/) page.
|
| 62 |
+
|
| 63 |
+
## Baseline
|
| 64 |
+
|
| 65 |
+
The [baseline](https://github.com/Audio-Editing-Challenge/Audio-Editing-Challenge-Baseline) is the AuK base model with the Prompt Enhancer disabled. It was scored on this test set with the Qwen3-Omni-30B-A3B-Instruct judge:
|
| 66 |
+
|
| 67 |
+
| IFR | CR | EMR |
|
| 68 |
+
| :---: | :---: | :---: |
|
| 69 |
+
| 40.33% | 74.70% | 7.60% |
|
| 70 |
+
|
| 71 |
+
Teams are ranked by EMR. Ties are broken by IFR, then by CR.
|
| 72 |
+
|
| 73 |
+
## Contact
|
| 74 |
+
|
| 75 |
+
zhikangniu@sjtu.edu.cn, tuwenming@sjtu.edu.cn
|