Datasets:
Size:
10K<n<100K
License:
MuGeminorum
commited on
Commit
•
7ea4927
1
Parent(s):
7911586
sync ms
Browse files
README.md
CHANGED
@@ -14,12 +14,15 @@ size_categories:
|
|
14 |
- 10K<n<100K
|
15 |
viewer: false
|
16 |
---
|
|
|
17 |
# Dataset Card for Music Genre
|
|
|
|
|
18 |
## Dataset Description
|
19 |
- **Homepage:** <https://ccmusic-database.github.io>
|
20 |
- **Repository:** <https://huggingface.co/datasets/ccmusic-database/music_genre>
|
21 |
- **Paper:** <https://doi.org/10.5281/zenodo.5676893>
|
22 |
-
- **Leaderboard:** <https://
|
23 |
- **Point of Contact:** <https://huggingface.co/ccmusic-database/music_genre>
|
24 |
|
25 |
### Dataset Summary
|
@@ -34,23 +37,42 @@ Multilingual
|
|
34 |
## Maintenance
|
35 |
```bash
|
36 |
GIT_LFS_SKIP_SMUDGE=1 git clone git@hf.co:datasets/ccmusic-database/music_genre
|
|
|
37 |
```
|
38 |
|
39 |
## Usage
|
40 |
-
|
41 |
```python
|
42 |
from datasets import load_dataset
|
43 |
|
44 |
-
dataset = load_dataset("ccmusic-database/music_genre")
|
|
|
|
|
45 |
|
46 |
-
for item in
|
47 |
print(item)
|
48 |
|
49 |
-
for item in
|
50 |
print(item)
|
51 |
```
|
52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
53 |
## Dataset Structure
|
|
|
54 |
<style>
|
55 |
#genres td {
|
56 |
vertical-align: middle !important;
|
@@ -62,17 +84,17 @@ for item in dataset["test"]:
|
|
62 |
</style>
|
63 |
<table id="genres">
|
64 |
<tr>
|
65 |
-
<td>mel(.jpg, 11.4s)</td>
|
66 |
-
<td>cqt(.jpg, 11.4s)</td>
|
67 |
-
<td>chroma(.jpg, 11.4s)</td>
|
68 |
<td>fst_level_label(2-class)</td>
|
69 |
<td>sec_level_label(9-class)</td>
|
70 |
<td>thr_level_label(16-class)</td>
|
71 |
</tr>
|
72 |
<tr>
|
73 |
-
<td><img src="
|
74 |
-
<td><img src="
|
75 |
-
<td><img src="
|
76 |
<td>1_Classic / 2_Non_classic</td>
|
77 |
<td>3_Symphony / 4_Opera / 5_Solo / 6_Chamber / 7_Pop / 8_Dance_and_house / 9_Indie / 10_Soul_or_r_and_b / 11_Rock</td>
|
78 |
<td>3_Symphony / 4_Opera / 5_Solo / 6_Chamber / 12_Pop_vocal_ballad / 13_Adult_contemporary / 14_Teen_pop / 15_Contemporary_dance_pop / 16_Dance_pop / 17_Classic_indie_pop / 18_Chamber_cabaret_and_art_pop / 10_Soul_or_r_and_b / 19_Adult_alternative_rock / 20_Uplifting_anthemic_rock / 21_Soft_rock / 22_Acoustic_pop</td>
|
@@ -87,6 +109,31 @@ for item in dataset["test"]:
|
|
87 |
</tr>
|
88 |
</table>
|
89 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
90 |
### Data Instances
|
91 |
.zip(.jpg)
|
92 |
|
@@ -122,11 +169,12 @@ for item in dataset["test"]:
|
|
122 |
```
|
123 |
|
124 |
### Data Splits
|
125 |
-
|
|
126 |
-
| :-------------: | :---: |
|
127 |
-
|
|
128 |
-
|
|
129 |
-
|
|
|
|
130 |
|
131 |
## Dataset Creation
|
132 |
### Curation Rationale
|
@@ -137,7 +185,7 @@ Promoting the development of AI in the music industry
|
|
137 |
Zhaorui Liu, Monan Zhou
|
138 |
|
139 |
#### Who are the source language producers?
|
140 |
-
Composers of the songs in dataset
|
141 |
|
142 |
### Annotations
|
143 |
#### Annotation process
|
@@ -192,16 +240,15 @@ SOFTWARE.
|
|
192 |
```
|
193 |
|
194 |
### Citation Information
|
195 |
-
```
|
196 |
@dataset{zhaorui_liu_2021_5676893,
|
197 |
-
author = {
|
198 |
-
title = {
|
199 |
-
month = {
|
200 |
-
year = {
|
201 |
-
publisher = {
|
202 |
-
version = {1.
|
203 |
-
|
204 |
-
url = {https://doi.org/10.5281/zenodo.5676893}
|
205 |
}
|
206 |
```
|
207 |
|
|
|
14 |
- 10K<n<100K
|
15 |
viewer: false
|
16 |
---
|
17 |
+
|
18 |
# Dataset Card for Music Genre
|
19 |
+
The raw dataset comprises approximately 1,700 musical pieces in .mp3 format, sourced from the NetEase music. The lengths of these pieces range from 270 to 300 seconds. All are sampled at the rate of 22,050 Hz. As the website providing the audio music includes style labels for the downloaded music, there are no specific annotators involved. Validation is achieved concurrently with the downloading process. They are categorized into a total of 16 genres.
|
20 |
+
|
21 |
## Dataset Description
|
22 |
- **Homepage:** <https://ccmusic-database.github.io>
|
23 |
- **Repository:** <https://huggingface.co/datasets/ccmusic-database/music_genre>
|
24 |
- **Paper:** <https://doi.org/10.5281/zenodo.5676893>
|
25 |
+
- **Leaderboard:** <https://www.modelscope.cn/datasets/ccmusic/music_genre>
|
26 |
- **Point of Contact:** <https://huggingface.co/ccmusic-database/music_genre>
|
27 |
|
28 |
### Dataset Summary
|
|
|
37 |
## Maintenance
|
38 |
```bash
|
39 |
GIT_LFS_SKIP_SMUDGE=1 git clone git@hf.co:datasets/ccmusic-database/music_genre
|
40 |
+
cd music_genre
|
41 |
```
|
42 |
|
43 |
## Usage
|
44 |
+
### Eval Subset
|
45 |
```python
|
46 |
from datasets import load_dataset
|
47 |
|
48 |
+
dataset = load_dataset("ccmusic-database/music_genre", name="eval")
|
49 |
+
for item in ds["train"]:
|
50 |
+
print(item)
|
51 |
|
52 |
+
for item in ds["validation"]:
|
53 |
print(item)
|
54 |
|
55 |
+
for item in ds["test"]:
|
56 |
print(item)
|
57 |
```
|
58 |
|
59 |
+
### Raw Subset
|
60 |
+
```python
|
61 |
+
from datasets import load_dataset
|
62 |
+
|
63 |
+
dataset = load_dataset("ccmusic-database/music_genre", name="default")
|
64 |
+
for item in ds["train"]:
|
65 |
+
print(item)
|
66 |
+
|
67 |
+
for item in ds["validation"]:
|
68 |
+
print(item)
|
69 |
+
|
70 |
+
for item in ds["test"]:
|
71 |
+
print(item)
|
72 |
+
```
|
73 |
+
|
74 |
## Dataset Structure
|
75 |
+
### Eval Subset
|
76 |
<style>
|
77 |
#genres td {
|
78 |
vertical-align: middle !important;
|
|
|
84 |
</style>
|
85 |
<table id="genres">
|
86 |
<tr>
|
87 |
+
<td>mel(.jpg, 11.4s, 48000Hz)</td>
|
88 |
+
<td>cqt(.jpg, 11.4s, 48000Hz)</td>
|
89 |
+
<td>chroma(.jpg, 11.4s, 48000Hz)</td>
|
90 |
<td>fst_level_label(2-class)</td>
|
91 |
<td>sec_level_label(9-class)</td>
|
92 |
<td>thr_level_label(16-class)</td>
|
93 |
</tr>
|
94 |
<tr>
|
95 |
+
<td><img src="./data/PqdpQP__ls-xo6lz93Q4y.jpeg"></td>
|
96 |
+
<td><img src="./data/EZfYLng40hh_FUudB9vvx.jpeg"></td>
|
97 |
+
<td><img src="./data/zviZ-rEKAvBCVFvKFml4R.jpeg"></td>
|
98 |
<td>1_Classic / 2_Non_classic</td>
|
99 |
<td>3_Symphony / 4_Opera / 5_Solo / 6_Chamber / 7_Pop / 8_Dance_and_house / 9_Indie / 10_Soul_or_r_and_b / 11_Rock</td>
|
100 |
<td>3_Symphony / 4_Opera / 5_Solo / 6_Chamber / 12_Pop_vocal_ballad / 13_Adult_contemporary / 14_Teen_pop / 15_Contemporary_dance_pop / 16_Dance_pop / 17_Classic_indie_pop / 18_Chamber_cabaret_and_art_pop / 10_Soul_or_r_and_b / 19_Adult_alternative_rock / 20_Uplifting_anthemic_rock / 21_Soft_rock / 22_Acoustic_pop</td>
|
|
|
109 |
</tr>
|
110 |
</table>
|
111 |
|
112 |
+
### Raw Subset
|
113 |
+
<table>
|
114 |
+
<tr>
|
115 |
+
<th>audio(.wav, 22050Hz)</th>
|
116 |
+
<th>mel(spectrogram, .jpg, 22050Hz)</th>
|
117 |
+
<th>fst_level_label(2-class)</th>
|
118 |
+
<th>sec_level_label(9-class)</th>
|
119 |
+
<th>thr_level_label(16-class)</th>
|
120 |
+
</tr>
|
121 |
+
<tr>
|
122 |
+
<td><audio controls src="./data/8bb58041d6b9d35db688bcedfde0fe39.mp3"></audio></td>
|
123 |
+
<td><img src="./data/8bb58041d6b9d35db688bcedfde0fe39.jpg"></td>
|
124 |
+
<td>1_Classic / 2_Non_classic</td>
|
125 |
+
<td>3_Symphony / 4_Opera / 5_Solo / 6_Chamber / 7_Pop / 8_Dance_and_house / 9_Indie / 10_Soul_or_r_and_b / 11_Rock</td>
|
126 |
+
<td>3_Symphony / 4_Opera / 5_Solo / 6_Chamber / 12_Pop_vocal_ballad / 13_Adult_contemporary / 14_Teen_pop / 15_Contemporary_dance_pop / 16_Dance_pop / 17_Classic_indie_pop / 18_Chamber_cabaret_and_art_pop / 10_Soul_or_r_and_b / 19_Adult_alternative_rock / 20_Uplifting_anthemic_rock / 21_Soft_rock / 22_Acoustic_pop</td>
|
127 |
+
</tr>
|
128 |
+
<tr>
|
129 |
+
<td>...</td>
|
130 |
+
<td>...</td>
|
131 |
+
<td>...</td>
|
132 |
+
<td>...</td>
|
133 |
+
<td>...</td>
|
134 |
+
</tr>
|
135 |
+
</table>
|
136 |
+
|
137 |
### Data Instances
|
138 |
.zip(.jpg)
|
139 |
|
|
|
169 |
```
|
170 |
|
171 |
### Data Splits
|
172 |
+
| Split | Eval | Raw |
|
173 |
+
| :-------------: | :---: | :---: |
|
174 |
+
| total | 36375 | 1713 |
|
175 |
+
| train(80%) | 29100 | 1370 |
|
176 |
+
| validation(10%) | 3637 | 171 |
|
177 |
+
| test(10%) | 3638 | 172 |
|
178 |
|
179 |
## Dataset Creation
|
180 |
### Curation Rationale
|
|
|
185 |
Zhaorui Liu, Monan Zhou
|
186 |
|
187 |
#### Who are the source language producers?
|
188 |
+
Composers of the songs in the dataset
|
189 |
|
190 |
### Annotations
|
191 |
#### Annotation process
|
|
|
240 |
```
|
241 |
|
242 |
### Citation Information
|
243 |
+
```bibtex
|
244 |
@dataset{zhaorui_liu_2021_5676893,
|
245 |
+
author = {Monan Zhou, Shenyang Xu, Zhaorui Liu, Zhaowen Wang, Feng Yu, Wei Li and Baoqiang Han},
|
246 |
+
title = {CCMusic: an Open and Diverse Database for Chinese and General Music Information Retrieval Research},
|
247 |
+
month = {mar},
|
248 |
+
year = {2024},
|
249 |
+
publisher = {HuggingFace},
|
250 |
+
version = {1.2},
|
251 |
+
url = {https://huggingface.co/ccmusic-database}
|
|
|
252 |
}
|
253 |
```
|
254 |
|
data/{genre_data.zip → 8bb58041d6b9d35db688bcedfde0fe39.jpg}
RENAMED
File without changes
|
data/8bb58041d6b9d35db688bcedfde0fe39.mp3
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d8406e26bbd885e74e0e6775d5150863d030a8fb9133063df503f4481dfa7687
|
3 |
+
size 4089932
|
data/EZfYLng40hh_FUudB9vvx.jpeg
ADDED
Git LFS Details
|
data/PqdpQP__ls-xo6lz93Q4y.jpeg
ADDED
Git LFS Details
|
data/zviZ-rEKAvBCVFvKFml4R.jpeg
ADDED
Git LFS Details
|
music_genre.py
CHANGED
@@ -1,13 +1,10 @@
|
|
|
|
1 |
import os
|
2 |
-
import socket
|
3 |
import random
|
4 |
import datasets
|
5 |
from datasets.tasks import ImageClassification
|
6 |
|
7 |
-
_NAMES_1 = {
|
8 |
-
1: "Classic",
|
9 |
-
2: "Non_classic"
|
10 |
-
}
|
11 |
|
12 |
_NAMES_2 = {
|
13 |
3: "Symphony",
|
@@ -18,7 +15,7 @@ _NAMES_2 = {
|
|
18 |
8: "Dance_and_house",
|
19 |
9: "Indie",
|
20 |
10: "Soul_or_r_and_b",
|
21 |
-
11: "Rock"
|
22 |
}
|
23 |
|
24 |
_NAMES_3 = {
|
@@ -37,44 +34,93 @@ _NAMES_3 = {
|
|
37 |
19: "Adult_alternative_rock",
|
38 |
20: "Uplifting_anthemic_rock",
|
39 |
21: "Soft_rock",
|
40 |
-
22: "Acoustic_pop"
|
41 |
}
|
42 |
|
43 |
-
_DBNAME = os.path.basename(__file__).split(
|
|
|
|
|
44 |
|
45 |
-
|
46 |
|
47 |
_CITATION = """\
|
48 |
@dataset{zhaorui_liu_2021_5676893,
|
49 |
-
author = {
|
50 |
-
title = {
|
51 |
-
month = {
|
52 |
-
year = {
|
53 |
-
publisher = {
|
54 |
-
version = {1.
|
55 |
-
|
56 |
-
url = {https://doi.org/10.5281/zenodo.5676893}
|
57 |
}
|
58 |
"""
|
59 |
|
60 |
_DESCRIPTION = """\
|
61 |
-
|
|
|
|
|
62 |
"""
|
63 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
64 |
|
65 |
class music_genre(datasets.GeneratorBasedBuilder):
|
66 |
-
|
67 |
-
|
|
|
|
|
68 |
features=datasets.Features(
|
69 |
{
|
70 |
"mel": datasets.Image(),
|
71 |
"cqt": datasets.Image(),
|
72 |
"chroma": datasets.Image(),
|
73 |
-
"fst_level_label": datasets.features.ClassLabel(
|
74 |
-
|
75 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
}
|
77 |
),
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
78 |
supervised_keys=("mel", "sec_level_label"),
|
79 |
homepage=_HOMEPAGE,
|
80 |
license="mit",
|
@@ -86,29 +132,58 @@ class music_genre(datasets.GeneratorBasedBuilder):
|
|
86 |
image_column="mel",
|
87 |
label_column="sec_level_label",
|
88 |
)
|
89 |
-
]
|
90 |
)
|
91 |
|
92 |
-
def
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
|
100 |
def _split_generators(self, dl_manager):
|
101 |
-
data_files = dl_manager.download_and_extract(self._cdn_url())
|
102 |
-
files = dl_manager.iter_files([data_files])
|
103 |
-
|
104 |
dataset = []
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
|
113 |
random.shuffle(dataset)
|
114 |
data_count = len(dataset)
|
@@ -118,49 +193,54 @@ class music_genre(datasets.GeneratorBasedBuilder):
|
|
118 |
return [
|
119 |
datasets.SplitGenerator(
|
120 |
name=datasets.Split.TRAIN,
|
121 |
-
gen_kwargs={
|
122 |
-
"files": dataset[:p80]
|
123 |
-
},
|
124 |
),
|
125 |
datasets.SplitGenerator(
|
126 |
name=datasets.Split.VALIDATION,
|
127 |
-
gen_kwargs={
|
128 |
-
"files": dataset[p80:p90]
|
129 |
-
},
|
130 |
),
|
131 |
datasets.SplitGenerator(
|
132 |
name=datasets.Split.TEST,
|
133 |
-
gen_kwargs={
|
134 |
-
"files": dataset[p90:]
|
135 |
-
},
|
136 |
),
|
137 |
]
|
138 |
|
139 |
-
def _calc_label(self, path, depth, substr=
|
140 |
spect = substr
|
141 |
-
dirpath = os.path.dirname(path)
|
142 |
substr_index = dirpath.find(spect)
|
143 |
if substr_index < 0:
|
144 |
-
spect = spect.replace(
|
145 |
substr_index = dirpath.find(spect)
|
146 |
|
147 |
-
labstr = dirpath[substr_index + len(spect):]
|
148 |
-
labs = labstr.split(
|
149 |
if len(labs) < 2:
|
150 |
-
labs = labstr.split(
|
151 |
|
152 |
if depth <= len(labs):
|
153 |
-
return int(labs[depth - 1].split(
|
154 |
else:
|
155 |
-
return int(labs[-1].split(
|
156 |
|
157 |
def _generate_examples(self, files):
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# import hashlib
|
2 |
import os
|
|
|
3 |
import random
|
4 |
import datasets
|
5 |
from datasets.tasks import ImageClassification
|
6 |
|
7 |
+
_NAMES_1 = {1: "Classic", 2: "Non_classic"}
|
|
|
|
|
|
|
8 |
|
9 |
_NAMES_2 = {
|
10 |
3: "Symphony",
|
|
|
15 |
8: "Dance_and_house",
|
16 |
9: "Indie",
|
17 |
10: "Soul_or_r_and_b",
|
18 |
+
11: "Rock",
|
19 |
}
|
20 |
|
21 |
_NAMES_3 = {
|
|
|
34 |
19: "Adult_alternative_rock",
|
35 |
20: "Uplifting_anthemic_rock",
|
36 |
21: "Soft_rock",
|
37 |
+
22: "Acoustic_pop",
|
38 |
}
|
39 |
|
40 |
+
_DBNAME = os.path.basename(__file__).split(".")[0]
|
41 |
+
|
42 |
+
_HOMEPAGE = f"https://www.modelscope.cn/datasets/ccmusic/{_DBNAME}"
|
43 |
|
44 |
+
_DOMAIN = f"https://www.modelscope.cn/api/v1/datasets/ccmusic/{_DBNAME}/repo?Revision=master&FilePath=data"
|
45 |
|
46 |
_CITATION = """\
|
47 |
@dataset{zhaorui_liu_2021_5676893,
|
48 |
+
author = {Monan Zhou, Shenyang Xu, Zhaorui Liu, Zhaowen Wang, Feng Yu, Wei Li and Zijin Li},
|
49 |
+
title = {CCMusic: an Open and Diverse Database for Chinese and General Music Information Retrieval Research},
|
50 |
+
month = {mar},
|
51 |
+
year = {2024},
|
52 |
+
publisher = {HuggingFace},
|
53 |
+
version = {1.2},
|
54 |
+
url = {https://huggingface.co/ccmusic-database}
|
|
|
55 |
}
|
56 |
"""
|
57 |
|
58 |
_DESCRIPTION = """\
|
59 |
+
The raw dataset comprises approximately 1,700 musical pieces in .mp3 format, sourced from the NetEase music. The lengths of these pieces range from 270 to 300 seconds. All are sampled at the rate of 48,000 Hz. As the website providing the audio music includes style labels for the downloaded music, there are no specific annotators involved. Validation is achieved concurrently with the downloading process. They are categorized into a total of 16 genres.
|
60 |
+
|
61 |
+
For the pre-processed version, audio is cut into an 11.4-second segment, resulting in 36,375 files, which are then transformed into Mel, CQT and Chroma. In the end, the data entry has six columns: the first three columns represent the Mel, CQT, and Chroma spectrogram slices in .jpg format, respectively, while the last three columns contain the labels for the three levels. The first level comprises two categories, the second level consists of nine categories, and the third level encompasses 16 categories. The entire dataset is shuffled and split into training, validation, and test sets in a ratio of 8:1:1. This dataset can be used for genre classification.
|
62 |
"""
|
63 |
|
64 |
+
_URLS = {
|
65 |
+
"audio": f"{_DOMAIN}/audio.zip",
|
66 |
+
"mel": f"{_DOMAIN}/mel.zip",
|
67 |
+
"eval": f"{_DOMAIN}/eval.zip",
|
68 |
+
}
|
69 |
+
|
70 |
+
|
71 |
+
class music_genre_Config(datasets.BuilderConfig):
|
72 |
+
def __init__(self, features, **kwargs):
|
73 |
+
super(music_genre_Config, self).__init__(
|
74 |
+
version=datasets.Version("1.2.0"), **kwargs
|
75 |
+
)
|
76 |
+
self.features = features
|
77 |
+
|
78 |
|
79 |
class music_genre(datasets.GeneratorBasedBuilder):
|
80 |
+
VERSION = datasets.Version("1.2.0")
|
81 |
+
BUILDER_CONFIGS = [
|
82 |
+
music_genre_Config(
|
83 |
+
name="eval",
|
84 |
features=datasets.Features(
|
85 |
{
|
86 |
"mel": datasets.Image(),
|
87 |
"cqt": datasets.Image(),
|
88 |
"chroma": datasets.Image(),
|
89 |
+
"fst_level_label": datasets.features.ClassLabel(
|
90 |
+
names=list(_NAMES_1.values())
|
91 |
+
),
|
92 |
+
"sec_level_label": datasets.features.ClassLabel(
|
93 |
+
names=list(_NAMES_2.values())
|
94 |
+
),
|
95 |
+
"thr_level_label": datasets.features.ClassLabel(
|
96 |
+
names=list(_NAMES_3.values())
|
97 |
+
),
|
98 |
}
|
99 |
),
|
100 |
+
),
|
101 |
+
music_genre_Config(
|
102 |
+
name="default",
|
103 |
+
features=datasets.Features(
|
104 |
+
{
|
105 |
+
"audio": datasets.Audio(sampling_rate=22050),
|
106 |
+
"mel": datasets.Image(),
|
107 |
+
"fst_level_label": datasets.features.ClassLabel(
|
108 |
+
names=list(_NAMES_1.values())
|
109 |
+
),
|
110 |
+
"sec_level_label": datasets.features.ClassLabel(
|
111 |
+
names=list(_NAMES_2.values())
|
112 |
+
),
|
113 |
+
"thr_level_label": datasets.features.ClassLabel(
|
114 |
+
names=list(_NAMES_3.values())
|
115 |
+
),
|
116 |
+
}
|
117 |
+
),
|
118 |
+
),
|
119 |
+
]
|
120 |
+
|
121 |
+
def _info(self):
|
122 |
+
return datasets.DatasetInfo(
|
123 |
+
features=self.config.features,
|
124 |
supervised_keys=("mel", "sec_level_label"),
|
125 |
homepage=_HOMEPAGE,
|
126 |
license="mit",
|
|
|
132 |
image_column="mel",
|
133 |
label_column="sec_level_label",
|
134 |
)
|
135 |
+
],
|
136 |
)
|
137 |
|
138 |
+
# def _str2md5(self, original_string):
|
139 |
+
# """
|
140 |
+
# Calculate and return the MD5 hash of a given string.
|
141 |
+
# Parameters:
|
142 |
+
# original_string (str): The original string for which the MD5 hash is to be computed.
|
143 |
+
# Returns:
|
144 |
+
# str: The hexadecimal representation of the MD5 hash.
|
145 |
+
# """
|
146 |
+
# # Create an md5 object
|
147 |
+
# md5_obj = hashlib.md5()
|
148 |
+
# # Update the md5 object with the original string encoded as bytes
|
149 |
+
# md5_obj.update(original_string.encode("utf-8"))
|
150 |
+
# # Retrieve the hexadecimal representation of the MD5 hash
|
151 |
+
# md5_hash = md5_obj.hexdigest()
|
152 |
+
# return md5_hash
|
153 |
|
154 |
def _split_generators(self, dl_manager):
|
|
|
|
|
|
|
155 |
dataset = []
|
156 |
+
if self.config.name == "eval":
|
157 |
+
data_files = dl_manager.download_and_extract(_URLS["eval"])
|
158 |
+
for path in dl_manager.iter_files([data_files]):
|
159 |
+
if os.path.basename(path).endswith(".jpg") and "mel" in path:
|
160 |
+
dataset.append(
|
161 |
+
{
|
162 |
+
"mel": path,
|
163 |
+
"cqt": path.replace("\\mel\\", "\\cqt\\").replace(
|
164 |
+
"/mel/", "/cqt/"
|
165 |
+
),
|
166 |
+
"chroma": path.replace("\\mel\\", "\\chroma\\").replace(
|
167 |
+
"/mel/", "/chroma/"
|
168 |
+
),
|
169 |
+
}
|
170 |
+
)
|
171 |
+
|
172 |
+
else:
|
173 |
+
files = {}
|
174 |
+
audio_files = dl_manager.download_and_extract(_URLS["audio"])
|
175 |
+
mel_files = dl_manager.download_and_extract(_URLS["mel"])
|
176 |
+
for path in dl_manager.iter_files([audio_files]):
|
177 |
+
fname: str = os.path.basename(path)
|
178 |
+
if fname.endswith(".mp3"):
|
179 |
+
files[fname.split(".mp")[0]] = {"audio": path}
|
180 |
+
|
181 |
+
for path in dl_manager.iter_files([mel_files]):
|
182 |
+
fname: str = os.path.basename(path)
|
183 |
+
if fname.endswith(".jpg"):
|
184 |
+
files[fname.split(".jp")[0]]["mel"] = path
|
185 |
+
|
186 |
+
dataset = list(files.values())
|
187 |
|
188 |
random.shuffle(dataset)
|
189 |
data_count = len(dataset)
|
|
|
193 |
return [
|
194 |
datasets.SplitGenerator(
|
195 |
name=datasets.Split.TRAIN,
|
196 |
+
gen_kwargs={"files": dataset[:p80]},
|
|
|
|
|
197 |
),
|
198 |
datasets.SplitGenerator(
|
199 |
name=datasets.Split.VALIDATION,
|
200 |
+
gen_kwargs={"files": dataset[p80:p90]},
|
|
|
|
|
201 |
),
|
202 |
datasets.SplitGenerator(
|
203 |
name=datasets.Split.TEST,
|
204 |
+
gen_kwargs={"files": dataset[p90:]},
|
|
|
|
|
205 |
),
|
206 |
]
|
207 |
|
208 |
+
def _calc_label(self, path, depth, substr="/mel/"):
|
209 |
spect = substr
|
210 |
+
dirpath: str = os.path.dirname(path)
|
211 |
substr_index = dirpath.find(spect)
|
212 |
if substr_index < 0:
|
213 |
+
spect = spect.replace("/", "\\")
|
214 |
substr_index = dirpath.find(spect)
|
215 |
|
216 |
+
labstr: str = dirpath[substr_index + len(spect) :]
|
217 |
+
labs = labstr.split("/")
|
218 |
if len(labs) < 2:
|
219 |
+
labs = labstr.split("\\")
|
220 |
|
221 |
if depth <= len(labs):
|
222 |
+
return int(labs[depth - 1].split("_")[0])
|
223 |
else:
|
224 |
+
return int(labs[-1].split("_")[0])
|
225 |
|
226 |
def _generate_examples(self, files):
|
227 |
+
if self.config.name == "eval":
|
228 |
+
for i, path in enumerate(files):
|
229 |
+
yield i, {
|
230 |
+
"mel": path["mel"],
|
231 |
+
"cqt": path["cqt"],
|
232 |
+
"chroma": path["chroma"],
|
233 |
+
"fst_level_label": _NAMES_1[self._calc_label(path["mel"], 1)],
|
234 |
+
"sec_level_label": _NAMES_2[self._calc_label(path["mel"], 2)],
|
235 |
+
"thr_level_label": _NAMES_3[self._calc_label(path["mel"], 3)],
|
236 |
+
}
|
237 |
+
|
238 |
+
else:
|
239 |
+
for i, path in enumerate(files):
|
240 |
+
yield i, {
|
241 |
+
"audio": path["audio"],
|
242 |
+
"mel": path["mel"],
|
243 |
+
"fst_level_label": _NAMES_1[self._calc_label(path["mel"], 1)],
|
244 |
+
"sec_level_label": _NAMES_2[self._calc_label(path["mel"], 2)],
|
245 |
+
"thr_level_label": _NAMES_3[self._calc_label(path["mel"], 3)],
|
246 |
+
}
|