For quick start, Just donwload preprocessed dataset from zenodo and check ecals_annotation folder
We introduce the extended tag version of CALS split (cleaned and artist-level stratified) for the Million Song Dataset (MSD). Different from the previously CALS dataset split, we provide 1054 vocabulary and caption level tag sequences instead of 50 small vocabs. Since we inherit the existing cals split, there is no difference in the test dataset. However, we use all tag annotations in the existing tag annotations msd50, msd500, and allmusic. This is the dataset repository for the paper: Toward Universal Text-to-Music Retrieval
Dataset is key-value type. The key is msdid, and value is python dictionary (item). Item consists of tag annotation data, and artist name, track title, and year meta data.
{
"TRSAGNY128F425391E":
{
'tag':
['aggressive', 'confrontational', 'energetic', 'alternative indie rock', 'self conscious', 'rowdy', 'bravado', 'pop rock', 'hardcore punk', 'passionate', 'confident', 'gutsy', 'swaggering', 'earnest', 'urgent', 'anguished distraught', 'straight edge', 'cathartic', 'punk', 'brash', 'rebellious', 'dramatic', 'alternative pop rock', 'street smart', 'summery', 'knotty', 'volatile', 'fiery', 'punk new wave', 'angry'],
'release': 'Sink With Kalifornija',
'artist_name': 'Youth Brigade',
'year': 1984,
'title': 'What Are You Fighting For?',
'track_id': 'TRSAGNY128F425391E'
}
}
If you want to use this dataset for audio-language representation learning or captioning, please refer to the this repository.
class ECALS_Dataset(Dataset):
"""
data_path (str): location of msu-benchmark
split (str): one of {TRAIN, VALID, TEST}
sr (int): sampling rate of waveform - 16000
num_chunks (int): chunk size of inference audio
"""
def __init__(self, data_path, split, sr, duration, num_chunks):
self.data_path = data_path
self.split = split
self.sr = sr
self.input_length = int(sr * duration)
self.num_chunks = num_chunks
self.msd_to_id = pickle.load(open(os.path.join(data_path, "lastfm_annotation", "MSD_id_to_7D_id.pkl"), 'rb'))
self.id_to_path = pickle.load(open(os.path.join(data_path, "lastfm_annotation", "7D_id_to_path.pkl"), 'rb'))
self.get_split()
self.get_file_list()
def get_split(self):
track_split = json.load(open(os.path.join(self.data_path, "ecals_annotation", "ecals_track_split.json"), "r"))
self.train_track = track_split['train_track'] + track_split['extra_track']
self.valid_track = track_split['valid_track']
self.test_track = track_split['test_track']
def get_file_list(self):
annotation = json.load(open(os.path.join(self.data_path, "ecals_annotation", "annotation.json"), 'r'))
self.list_of_label = json.load(open(os.path.join(self.data_path, "ecals_annotation", "ecals_tags.json"), 'r'))
self.tag_to_idx = {i:idx for idx, i in enumerate(self.list_of_label)}
if self.split == "TRAIN":
self.fl = [annotation[i] for i in self.train_track]
elif self.split == "VALID":
self.fl = [annotation[i] for i in self.valid_track]
elif self.split == "TEST":
self.fl = [annotation[i] for i in self.test_track]
else:
raise ValueError(f"Unexpected split name: {self.split}")
del annotation
def audio_load(self, msd_id):
audio_path = self.id_to_path[self.msd_to_id[msd_id]]
audio = np.load(os.path.join(self.data_path, "npy", audio_path.replace(".mp3",".npy")), mmap_mode='r')
random_idx = random.randint(0, audio.shape[-1]-self.input_length)
audio = torch.from_numpy(np.array(audio[random_idx:random_idx+self.input_length]))
return audio
def tag_to_binary(self, tag_list):
bainry = np.zeros([len(self.list_of_label),], dtype=np.float32)
for tag in tag_list:
bainry[self.tag_to_idx[tag]] = 1.0
return bainry
def __getitem__(self, index):
item = self.fl[index]
tag_list = item['tag']
binary = self.tag_to_binary(tag_list)
audio_tensor = self.audio_load(item['track_id'])
return {
"audio":audio_tensor,
"binary":binary,
"tag_list":tag_list
}
def __len__(self):
return len(self.fl)Train Track: 444865 (== CALS train + student)
Valid Track: 34481 (== CALS valid track)
Test Track: 34631 (== CALS test track)
Unique Tag: 1054
Unique Tag Caption: 139541
Unique Artist: 32650
Unique Album : 89920
Unique Year: 90
wget https://zenodo.org/record/7107130/files/dataset.tar.gz
tar -xvf dataset.tar.gz
cd dataset
wget http://millionsongdataset.com/sites/default/files/AdditionalFiles/track_metadata.db
└── dataset
track_metadata.db
├── allmusic_annotation
│ └── ground_truth_assignments
│ ├── AMG_Multilabel_tagsets
│ │ ├── msd_amglabels_all.h5
...
├── cals_annotation
│ ├── cals_error.npy
│ ├── cals_tags.npy
...
├── lastfm_annotation
│ ├── 50tagList.txt
│ ├── filtered_list_test.cP
│ ├── filtered_list_train.cP
│ ├── msd_id_to_tag_vector.cP
...
├── msd500_annotation
│ ├── dataset_stats.txt
│ ├── selected_tags.tsv
│ └── track_tags.tsv
└── ecals_annotation
├── annotation.json
├── ecals_tags.json
├── ecals_tag_stats.json
├── ecals_track_split.json
└── multiquery_samples.json
cd preprocessing
python main.py
Due to copyright issue, we don't provide audio data in this page.
Please consider citing our paper in your publications if the project helps your research. BibTeX reference is as follow.
@inproceedings{doh2023toward,
title={Toward Universal Text-to-Music Retrieval},
author={Doh, SeungHeon and Won, Minz and Choi, Keunwoo and Nam, Juhan},
booktitle={ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
year={2023}
}