liu-hanghang commited on
Commit
3f3db4a
·
verified ·
1 Parent(s): 5753cf7

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. README.md +11 -14
  2. metadata.jsonl +0 -0
README.md CHANGED
@@ -1,7 +1,3 @@
1
- ---
2
- language:
3
- - en
4
- ---
5
  # JAME Dataset
6
 
7
  **JAME** (JAM Evaluation) is a comprehensive music dataset containing 250 high-quality music tracks designed for **standardized evaluation of song generation models**.
@@ -18,10 +14,10 @@ arXiv:2507.20880 (2025)
18
 
19
  ## Dataset Overview
20
 
21
- - **Total Tracks**: 250 carefully curated tracks
22
  - **Duration**: Approximately 230 seconds per track
23
  - **Genres**: Hip-Hop/Rap, Rock/Metal, Electronic/Dance, R&B/Soul/Jazz, Country/Folk
24
- - **Release Time**: After 1st May 2025
25
  - **Purpose**: Standardized evaluation benchmark for song generation models
26
 
27
  ## Dataset Structure
@@ -62,8 +58,8 @@ Each line in `metadata.jsonl` contains a JSON object with the following fields:
62
  "duration": 180,
63
  "year": 2025,
64
  "genre": "Hip-Hop/Rap",
65
- "transcription_path": "eval_data/jame/transcriptions/Artist - Title.json",
66
- "struct_path": "eval_data/jame/struct/Artist - Title.json",
67
  "song_id": "spotify_track_id"
68
  }
69
  ```
@@ -84,12 +80,13 @@ Each line in `metadata.jsonl` contains a JSON object with the following fields:
84
 
85
  ## Genre Distribution
86
 
87
- The dataset contains tracks from the following genres:
88
- - **Hip-Hop/Rap**: Urban and rap music
89
- - **Rock/Metal**: Rock, alternative, and metal tracks
90
- - **Electronic/Dance**: Electronic, EDM, and dance music
91
- - **R&B/Soul/Jazz**: R&B, soul, and jazz-influenced tracks
92
- - **Country/Folk**: Country, folk, and acoustic music
 
93
 
94
  ## File Formats
95
 
 
 
 
 
 
1
  # JAME Dataset
2
 
3
  **JAME** (JAM Evaluation) is a comprehensive music dataset containing 250 high-quality music tracks designed for **standardized evaluation of song generation models**.
 
14
 
15
  ## Dataset Overview
16
 
17
+ - **Total Tracks**: 250 carefully curated tracks (50 per genre, 5 genres total)
18
  - **Duration**: Approximately 230 seconds per track
19
  - **Genres**: Hip-Hop/Rap, Rock/Metal, Electronic/Dance, R&B/Soul/Jazz, Country/Folk
20
+ - **Release Time**: select tracks released after 1st May 2025 to avoid data contaimination.
21
  - **Purpose**: Standardized evaluation benchmark for song generation models
22
 
23
  ## Dataset Structure
 
58
  "duration": 180,
59
  "year": 2025,
60
  "genre": "Hip-Hop/Rap",
61
+ "transcription_path": "transcriptions/Artist - Title.json",
62
+ "struct_path": "struct/Artist - Title.json",
63
  "song_id": "spotify_track_id"
64
  }
65
  ```
 
80
 
81
  ## Genre Distribution
82
 
83
+ The dataset contains 250 tracks evenly distributed across 5 genres (50 tracks per genre):
84
+
85
+ - **Hip-Hop/Rap** (50 tracks): Urban and rap music
86
+ - **Rock/Metal** (50 tracks): Rock, alternative, and metal tracks
87
+ - **Electronic/Dance** (50 tracks): Electronic, EDM, and dance music
88
+ - **R&B/Soul/Jazz** (50 tracks): R&B, soul, and jazz-influenced tracks
89
+ - **Country/Folk** (50 tracks): Country, folk, and acoustic music
90
 
91
  ## File Formats
92
 
metadata.jsonl CHANGED
The diff for this file is too large to render. See raw diff