yangwang825
commited on
Commit
•
7619d15
1
Parent(s):
39faa4d
Update gtzan.py
Browse files
gtzan.py
CHANGED
@@ -9,6 +9,103 @@ GENRES = [
|
|
9 |
"jazz", "metal", "pop", "reggae", "rock"
|
10 |
]
|
11 |
EXTENSION = '.au'
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
FILTERED_TEST = [
|
13 |
"blues.00012",
|
14 |
"blues.00013",
|
@@ -944,99 +1041,4 @@ FILTERED_VALID = [
|
|
944 |
"rock.00083",
|
945 |
"rock.00084",
|
946 |
"rock.00085",
|
947 |
-
]
|
948 |
-
|
949 |
-
|
950 |
-
class GTZAN(datasets.GeneratorBasedBuilder):
|
951 |
-
"""
|
952 |
-
The partitioning of GTZAN is "fault-filtered",
|
953 |
-
which is constructed by hand to include 443/197/290 excerpts [1].
|
954 |
-
|
955 |
-
References
|
956 |
-
----------
|
957 |
-
[1] https://arxiv.org/pdf/1507.04761.pdf
|
958 |
-
"""
|
959 |
-
def _info(self):
|
960 |
-
return datasets.DatasetInfo(
|
961 |
-
description='Music genre classification dataset',
|
962 |
-
features=datasets.Features(
|
963 |
-
{
|
964 |
-
"filepath": datasets.Value("string"),
|
965 |
-
"audio": datasets.Audio(sampling_rate=SAMPLE_RATE),
|
966 |
-
"genre": datasets.ClassLabel(names=GENRES),
|
967 |
-
}
|
968 |
-
),
|
969 |
-
)
|
970 |
-
|
971 |
-
def _split_generators(self, dl_manager):
|
972 |
-
data_dir = dl_manager.download_and_extract("genres.tar.gz")
|
973 |
-
return [
|
974 |
-
datasets.SplitGenerator(
|
975 |
-
name=datasets.Split.TRAIN,
|
976 |
-
gen_kwargs={
|
977 |
-
"data_dir": data_dir,
|
978 |
-
"split": 'train'
|
979 |
-
},
|
980 |
-
),
|
981 |
-
datasets.SplitGenerator(
|
982 |
-
name=datasets.Split.VALIDATION,
|
983 |
-
gen_kwargs={
|
984 |
-
"data_dir": data_dir,
|
985 |
-
"split": 'validation'
|
986 |
-
},
|
987 |
-
),
|
988 |
-
datasets.SplitGenerator(
|
989 |
-
name=datasets.Split.TEST,
|
990 |
-
gen_kwargs={
|
991 |
-
"data_dir": data_dir,
|
992 |
-
"split": 'test'
|
993 |
-
},
|
994 |
-
)
|
995 |
-
]
|
996 |
-
|
997 |
-
def _generate_examples(self, data_dir, split):
|
998 |
-
_, _walker = fast_scandir(data_dir, [EXTENSION], recursive=True)
|
999 |
-
if split == 'train':
|
1000 |
-
_walker = [fileid for fileid in _walker if Path(fileid).stem in FILTERED_TRAIN]
|
1001 |
-
elif split == 'validation':
|
1002 |
-
_walker = [fileid for fileid in _walker if Path(fileid).stem in FILTERED_VALID]
|
1003 |
-
elif split == 'test':
|
1004 |
-
_walker = [fileid for fileid in _walker if Path(fileid).stem in FILTERED_TEST]
|
1005 |
-
|
1006 |
-
for idx, fileid in enumerate(_walker):
|
1007 |
-
yield idx, {
|
1008 |
-
'filepath': fileid,
|
1009 |
-
'audio': fileid,
|
1010 |
-
'genre': default_find_classes(fileid)
|
1011 |
-
}
|
1012 |
-
|
1013 |
-
|
1014 |
-
def default_find_classes(audio_path):
|
1015 |
-
return Path(audio_path).parent.stem
|
1016 |
-
|
1017 |
-
|
1018 |
-
def fast_scandir(path: str, exts: tp.List[str], recursive: bool = False):
|
1019 |
-
# Scan files recursively faster than glob
|
1020 |
-
# From github.com/drscotthawley/aeiou/blob/main/aeiou/core.py
|
1021 |
-
subfolders, files = [], []
|
1022 |
-
|
1023 |
-
try: # hope to avoid 'permission denied' by this try
|
1024 |
-
for f in os.scandir(path):
|
1025 |
-
try: # 'hope to avoid too many levels of symbolic links' error
|
1026 |
-
if f.is_dir():
|
1027 |
-
subfolders.append(f.path)
|
1028 |
-
elif f.is_file():
|
1029 |
-
if os.path.splitext(f.name)[1].lower() in exts:
|
1030 |
-
files.append(f.path)
|
1031 |
-
except Exception:
|
1032 |
-
pass
|
1033 |
-
except Exception:
|
1034 |
-
pass
|
1035 |
-
|
1036 |
-
if recursive:
|
1037 |
-
for path in list(subfolders):
|
1038 |
-
sf, f = fast_scandir(path, exts, recursive=recursive)
|
1039 |
-
subfolders.extend(sf)
|
1040 |
-
files.extend(f) # type: ignore
|
1041 |
-
|
1042 |
-
return subfolders, files
|
|
|
9 |
"jazz", "metal", "pop", "reggae", "rock"
|
10 |
]
|
11 |
EXTENSION = '.au'
|
12 |
+
|
13 |
+
|
14 |
+
class GTZAN(datasets.GeneratorBasedBuilder):
|
15 |
+
"""
|
16 |
+
The partitioning of GTZAN is "fault-filtered",
|
17 |
+
which is constructed by hand to include 443/197/290 excerpts [1].
|
18 |
+
|
19 |
+
References
|
20 |
+
----------
|
21 |
+
[1] https://arxiv.org/pdf/1507.04761.pdf
|
22 |
+
"""
|
23 |
+
def _info(self):
|
24 |
+
return datasets.DatasetInfo(
|
25 |
+
description='Music genre classification dataset',
|
26 |
+
features=datasets.Features(
|
27 |
+
{
|
28 |
+
"filepath": datasets.Value("string"),
|
29 |
+
"audio": datasets.Audio(sampling_rate=SAMPLE_RATE),
|
30 |
+
"genre": datasets.ClassLabel(names=GENRES),
|
31 |
+
}
|
32 |
+
),
|
33 |
+
)
|
34 |
+
|
35 |
+
def _split_generators(self, dl_manager):
|
36 |
+
data_dir = dl_manager.download_and_extract("gtzan.tar.gz")
|
37 |
+
return [
|
38 |
+
datasets.SplitGenerator(
|
39 |
+
name=datasets.Split.TRAIN,
|
40 |
+
gen_kwargs={
|
41 |
+
"data_dir": data_dir,
|
42 |
+
"split": 'train'
|
43 |
+
},
|
44 |
+
),
|
45 |
+
datasets.SplitGenerator(
|
46 |
+
name=datasets.Split.VALIDATION,
|
47 |
+
gen_kwargs={
|
48 |
+
"data_dir": data_dir,
|
49 |
+
"split": 'validation'
|
50 |
+
},
|
51 |
+
),
|
52 |
+
datasets.SplitGenerator(
|
53 |
+
name=datasets.Split.TEST,
|
54 |
+
gen_kwargs={
|
55 |
+
"data_dir": data_dir,
|
56 |
+
"split": 'test'
|
57 |
+
},
|
58 |
+
)
|
59 |
+
]
|
60 |
+
|
61 |
+
def _generate_examples(self, data_dir, split):
|
62 |
+
_, _walker = fast_scandir(data_dir, [EXTENSION], recursive=True)
|
63 |
+
if split == 'train':
|
64 |
+
_walker = [fileid for fileid in _walker if Path(fileid).stem in FILTERED_TRAIN]
|
65 |
+
elif split == 'validation':
|
66 |
+
_walker = [fileid for fileid in _walker if Path(fileid).stem in FILTERED_VALID]
|
67 |
+
elif split == 'test':
|
68 |
+
_walker = [fileid for fileid in _walker if Path(fileid).stem in FILTERED_TEST]
|
69 |
+
|
70 |
+
for idx, fileid in enumerate(_walker):
|
71 |
+
yield idx, {
|
72 |
+
'filepath': fileid,
|
73 |
+
'audio': fileid,
|
74 |
+
'genre': default_find_classes(fileid)
|
75 |
+
}
|
76 |
+
|
77 |
+
|
78 |
+
def default_find_classes(audio_path):
|
79 |
+
return Path(audio_path).parent.stem
|
80 |
+
|
81 |
+
|
82 |
+
def fast_scandir(path: str, exts: tp.List[str], recursive: bool = False):
|
83 |
+
# Scan files recursively faster than glob
|
84 |
+
# From github.com/drscotthawley/aeiou/blob/main/aeiou/core.py
|
85 |
+
subfolders, files = [], []
|
86 |
+
|
87 |
+
try: # hope to avoid 'permission denied' by this try
|
88 |
+
for f in os.scandir(path):
|
89 |
+
try: # 'hope to avoid too many levels of symbolic links' error
|
90 |
+
if f.is_dir():
|
91 |
+
subfolders.append(f.path)
|
92 |
+
elif f.is_file():
|
93 |
+
if os.path.splitext(f.name)[1].lower() in exts:
|
94 |
+
files.append(f.path)
|
95 |
+
except Exception:
|
96 |
+
pass
|
97 |
+
except Exception:
|
98 |
+
pass
|
99 |
+
|
100 |
+
if recursive:
|
101 |
+
for path in list(subfolders):
|
102 |
+
sf, f = fast_scandir(path, exts, recursive=recursive)
|
103 |
+
subfolders.extend(sf)
|
104 |
+
files.extend(f) # type: ignore
|
105 |
+
|
106 |
+
return subfolders, files
|
107 |
+
|
108 |
+
|
109 |
FILTERED_TEST = [
|
110 |
"blues.00012",
|
111 |
"blues.00013",
|
|
|
1041 |
"rock.00083",
|
1042 |
"rock.00084",
|
1043 |
"rock.00085",
|
1044 |
+
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|