def get_results_file_name(boundaries_id, labels_id, config, ds_name):
"""Based on the config and the dataset, get the file name to store the
results."""
if ds_name == "*":
ds_name = "All"
utils.ensure_dir(msaf.results_dir)
file_name = os.path.join(msaf.results_dir, "results_%s" % ds_name)
file_name += "_boundsE%s_labelsE%s" % (boundaries_id, labels_id)
sorted_keys = sorted(config.keys(),
cmp=lambda x, y: cmp(x.lower(), y.lower()))
for key in sorted_keys:
file_name += "_%sE%s" % (key, str(config[key]).replace("/", "_"))
return file_name + msaf.results_ext
def get_results_file_name(boundaries_id, labels_id, config, annotator_id):
"""Based on the config and the dataset, get the file name to store the
results."""
utils.ensure_dir(msaf.config.results_dir)
file_name = os.path.join(msaf.config.results_dir, "results")
file_name += "_boundsE%s_labelsE%s" % (boundaries_id, labels_id)
file_name += "_annotatorE%d" % (annotator_id)
sorted_keys = sorted(config.keys(), key=str.lower)
for key in sorted_keys:
file_name += "_%sE%s" % (key, str(config[key]).replace("/", "_"))
# Check for max file length
if len(file_name) > 255 - len(msaf.config.results_ext):
file_name = file_name[:255 - len(msaf.config.results_ext)]
return file_name + msaf.config.results_ext
def get_results_file_name(boundaries_id, labels_id, config, ds_name,
annotator_id):
"""Based on the config and the dataset, get the file name to store the
results."""
utils.ensure_dir(msaf.config.results_dir)
file_name = os.path.join(msaf.config.results_dir, "results")
file_name += "_boundsE%s_labelsE%s" % (boundaries_id, labels_id)
file_name += "_annotatorE%d" % (annotator_id)
sorted_keys = sorted(config.keys(), key=str.lower)
for key in sorted_keys:
file_name += "_%sE%s" % (key, str(config[key]).replace("/", "_"))
# Check for max file length
if len(file_name) > 255 - len(msaf.config.results_ext):
file_name = file_name[:255 - len(msaf.config.results_ext)]
return file_name + msaf.config.results_ext
def process(in_path,
sonify_beats=False,
n_jobs=1,
overwrite=False,
out_file="out.json",
out_beats="out_beats.wav",
ds_name="*"):
"""Main process to compute features.
Parameters
----------
in_path: str
Path to the file or dataset to compute the features.
sonify_beats: bool
Whether to sonify the beats on top of the audio file
(single file mode only).
n_jobs: int
Number of threads (collection mode only).
overwrite: bool
Whether to overwrite the previously computed features.
out_file: str
Path to the output json file (single file mode only).
out_beats: str
Path to the new file containing the sonified beats.
ds_name: str
Name of the prefix of the dataset (e.g., Beatles)
"""
# If in_path it's a file, we only compute one file
if os.path.isfile(in_path):
file_struct = FileStruct(in_path)
file_struct.features_file = out_file
compute_all_features(file_struct, sonify_beats, overwrite, out_beats)
elif os.path.isdir(in_path):
# Check that in_path exists
utils.ensure_dir(in_path)
# Get files
file_structs = io.get_dataset_files(in_path, ds_name=ds_name)
# Compute features using joblib
Parallel(n_jobs=n_jobs)(delayed(compute_all_features)(
file_struct, sonify_beats, overwrite, out_beats)
for file_struct in file_structs)
def process(in_path, audio_beats=False, n_jobs=1, overwrite=False):
"""Main process."""
# If in_path it's a file, we only compute one file
if os.path.isfile(in_path):
compute_all_features(in_path, audio_beats, overwrite)
elif os.path.isdir(in_path):
# Check that in_path exists
utils.ensure_dir(in_path)
# Get files
file_structs = io.get_dataset_files(in_path)
# Compute features using joblib
Parallel(n_jobs=n_jobs)(delayed(compute_all_features)(
file_struct, audio_beats, overwrite)
for file_struct in file_structs)
def process(
in_path, sonify_beats=False, n_jobs=1, overwrite=False, out_file="out.json", out_beats="out_beats.wav", ds_name="*"
):
"""Main process to compute features.
Parameters
----------
in_path: str
Path to the file or dataset to compute the features.
sonify_beats: bool
Whether to sonify the beats on top of the audio file
(single file mode only).
n_jobs: int
Number of threads (collection mode only).
overwrite: bool
Whether to overwrite the previously computed features.
out_file: str
Path to the output json file (single file mode only).
out_beats: str
Path to the new file containing the sonified beats.
ds_name: str
Name of the prefix of the dataset (e.g., Beatles)
"""
# If in_path it's a file, we only compute one file
if os.path.isfile(in_path):
file_struct = FileStruct(in_path)
file_struct.features_file = out_file
compute_all_features(file_struct, sonify_beats, overwrite, out_beats)
elif os.path.isdir(in_path):
# Check that in_path exists
utils.ensure_dir(in_path)
# Get files
file_structs = io.get_dataset_files(in_path, ds_name=ds_name)
# Compute features using joblib
Parallel(n_jobs=n_jobs)(
delayed(compute_all_features)(file_struct, sonify_beats, overwrite, out_beats)
for file_struct in file_structs
)
def get_dataset_files(in_path, ds_name="*"):
"""Gets the files of the dataset with a prefix of ds_name."""
# All datasets
ds_dict = {
"Beatles" : "Isophonics",
"Cerulean" : "Cerulean",
"Epiphyte" : "Epiphyte",
"Isophonics": "Isophonics",
"SALAMI" : "SALAMI",
"SALAMI-i" : "SALAMI",
"*" : "*"
}
try:
prefix = ds_dict[ds_name]
except KeyError:
raise RuntimeError("Dataset %s is not valid. Valid datasets are: %s" %
(ds_name, ds_dict.keys()))
# Get audio files
audio_files = []
for ext in msaf.Dataset.audio_exts:
audio_files += glob.glob(os.path.join(in_path, msaf.Dataset.audio_dir,
("%s_*" + ext) % prefix))
# Check for datasets with different prefix
if len(audio_files) == 0:
for ext in msaf.Dataset.audio_exts:
audio_files += glob.glob(os.path.join(in_path,
msaf.Dataset.audio_dir,
"*" + ext))
# Make sure directories exist
utils.ensure_dir(os.path.join(in_path, msaf.Dataset.features_dir))
utils.ensure_dir(os.path.join(in_path, msaf.Dataset.estimations_dir))
utils.ensure_dir(os.path.join(in_path, msaf.Dataset.references_dir))
# Get the file structs
file_structs = []
for audio_file in audio_files:
file_structs.append(FileStruct(audio_file))
# Filter by the beatles
if ds_name == "Beatles":
file_structs = filter_by_artist(file_structs, "The Beatles")
# Salami Internet hack
if ds_name == "SALAMI-i":
file_structs = get_SALAMI_internet(file_structs)
# Sort by audio file name
file_structs = sorted(file_structs,
key=lambda file_struct: file_struct.audio_file)
return file_structs
def get_dataset_files(in_path, ds_name="*"):
"""Gets the files of the dataset with a prefix of ds_name."""
# All datasets
ds_dict = {
"Beatles": "Isophonics",
"Cerulean": "Cerulean",
"Epiphyte": "Epiphyte",
"Isophonics": "Isophonics",
"SALAMI": "SALAMI",
"SALAMI-i": "SALAMI",
"*": "*"
}
try:
prefix = ds_dict[ds_name]
except KeyError:
raise RuntimeError("Dataset %s is not valid. Valid datasets are: %s" %
(ds_name, ds_dict.keys()))
# Get audio files
audio_files = []
for ext in msaf.Dataset.audio_exts:
audio_files += glob.glob(
os.path.join(in_path, msaf.Dataset.audio_dir,
("%s_*" + ext) % prefix))
# Check for datasets with different prefix
if len(audio_files) == 0:
for ext in msaf.Dataset.audio_exts:
audio_files += glob.glob(
os.path.join(in_path, msaf.Dataset.audio_dir, "*" + ext))
# Make sure directories exist
utils.ensure_dir(os.path.join(in_path, msaf.Dataset.features_dir))
utils.ensure_dir(os.path.join(in_path, msaf.Dataset.estimations_dir))
utils.ensure_dir(os.path.join(in_path, msaf.Dataset.references_dir))
# Get the file structs
file_structs = []
for audio_file in audio_files:
file_structs.append(FileStruct(audio_file))
# Filter by the beatles
if ds_name == "Beatles":
file_structs = filter_by_artist(file_structs, "The Beatles")
# Salami Internet hack
if ds_name == "SALAMI-i":
file_structs = get_SALAMI_internet(file_structs)
# Sort by audio file name
file_structs = sorted(file_structs,
key=lambda file_struct: file_struct.audio_file)
return file_structs
def get_dataset_files(in_path, ds_name="*"):
"""Gets the files of the dataset with a prefix of ds_name."""
# All datasets
ds_dict = {
"Beatles" : "Isophonics",
"Cerulean" : "Cerulean",
"Epiphyte" : "Epiphyte",
"Isophonics": "Isophonics",
"SALAMI" : "SALAMI",
"SALAMI-i" : "SALAMI",
"CJ": "CJ",
"*" : "*"
}
try:
prefix = ds_dict[ds_name]
except KeyError:
raise RuntimeError("Dataset %s is not valid. Valid datasets are: %s" %
(ds_name, ds_dict.keys()))
audio_files = [x for x in os.listdir(os.path.join(in_path, msaf.Dataset.audio_dir)) if (not x.startswith('.') and (x.endswith('.aif') or \
x.endswith('.wav') or x.endswith('.mp3')) )]
audio_files = [os.path.join(in_path, msaf.Dataset.audio_dir, x) for x in audio_files]
# Make sure directories exist
utils.ensure_dir(os.path.join(in_path, msaf.Dataset.features_dir))
utils.ensure_dir(os.path.join(in_path, msaf.Dataset.estimations_dir))
utils.ensure_dir(os.path.join(in_path, msaf.Dataset.references_dir))
# Get the file structs
file_structs = []
for audio_file in audio_files:
file_structs.append(FileStruct(audio_file))
# Filter by the beatles
if ds_name == "Beatles":
file_structs = filter_by_artist(file_structs, "The Beatles")
# Salami Internet hack
if ds_name == "SALAMI-i":
file_structs = get_SALAMI_internet(file_structs)
# Sort by audio file name
file_structs = sorted(file_structs,
key=lambda file_struct: file_struct.audio_file)
return file_structs
def get_dataset_files(in_path):
"""Gets the files of the given dataset."""
# Get audio files
audio_files = []
for ext in ds_config.audio_exts:
audio_files += glob.glob(
os.path.join(in_path, ds_config.audio_dir, "*" + ext))
# Make sure directories exist
utils.ensure_dir(os.path.join(in_path, ds_config.features_dir))
utils.ensure_dir(os.path.join(in_path, ds_config.estimations_dir))
utils.ensure_dir(os.path.join(in_path, ds_config.references_dir))
# Get the file structs
file_structs = []
for audio_file in audio_files:
file_structs.append(FileStruct(audio_file))
# Sort by audio file name
file_structs = sorted(file_structs,
key=lambda file_struct: file_struct.audio_file)
return file_structs
def get_dataset_files(in_path):
"""Gets the files of the given dataset."""
# Get audio files
audio_files = []
for ext in ds_config.audio_exts:
audio_files += glob.glob(
os.path.join(in_path, ds_config.audio_dir, "*" + ext))
# Make sure directories exist
utils.ensure_dir(os.path.join(in_path, ds_config.features_dir))
utils.ensure_dir(os.path.join(in_path, ds_config.estimations_dir))
utils.ensure_dir(os.path.join(in_path, ds_config.references_dir))
# Get the file structs
file_structs = []
for audio_file in audio_files:
file_structs.append(FileStruct(audio_file))
# Sort by audio file name
file_structs = sorted(file_structs,
key=lambda file_struct: file_struct.audio_file)
return file_structs