if __name__ == '__main__':
# Load in list of MSD entries
msd_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'msd', 'index'))
with msd_index.searcher() as searcher:
msd_list = list(searcher.documents())
# Load in list of MSD entries
midi_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'clean_midi', 'index'))
with midi_index.searcher() as searcher:
midi_list = list(searcher.documents())
# Load MSD embeddings
msd_embedding_datas = experiment_utils.load_precomputed_data(
msd_list, os.path.join(RESULTS_PATH, 'pse_msd_embeddings'))
# Load in hash sequences (and metadata) for all MSD entries
msd_sequence_datas = experiment_utils.load_precomputed_data(
msd_list, os.path.join(RESULTS_PATH, 'dhs_msd_hash_sequences'))
# Ignore entries with very short sequences
msd_embedding_datas = [
e for d, e in zip(msd_sequence_datas, msd_embedding_datas)
if len(d['hash_sequence']) > MIN_SEQUENCE_LENGTH]
msd_sequence_datas = [d for d in msd_sequence_datas
if len(d['hash_sequence']) > MIN_SEQUENCE_LENGTH]
# Create a big matrix of the embeddings
msd_embeddings = np.concatenate(
[d['embedding'] for d in msd_embedding_datas], axis=0)
# Construct paths to each feature file
msd_feature_paths = [
os.path.join(DATA_PATH, 'msd', 'h5', d['path']) + '.h5'
if __name__ == '__main__':
# Load in list of MSD entries
msd_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'msd', 'index'))
with msd_index.searcher() as searcher:
msd_list = list(searcher.documents())
# Load in list of MSD entries
midi_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'clean_midi', 'index'))
with midi_index.searcher() as searcher:
midi_list = list(searcher.documents())
# Load in hash sequences (and metadata) for all MSD entries
msd_data = experiment_utils.load_precomputed_data(
msd_list, os.path.join(RESULTS_PATH, 'tpaa_msd_hash_sequences'))
# Ignore very short sequences
msd_data = [d for d in msd_data
if len(d['hash_sequence']) > MIN_SEQUENCE_LENGTH]
# Get a list of valid MIDI-MSD match pairs
midi_msd_mapping = experiment_utils.get_valid_matches(
os.path.join(RESULTS_PATH, '{}_pairs.csv'.format(SPLIT)),
SCORE_THRESHOLD,
os.path.join(RESULTS_PATH, 'clean_midi_aligned', 'h5'))
midi_datas, midi_index_mapping = experiment_utils.load_valid_midi_datas(
midi_msd_mapping, msd_data, midi_list,
os.path.join(RESULTS_PATH, 'tpaa_clean_midi_hash_sequences'))
# Run match_one_midi for each MIDI data and MSD index list
if __name__ == '__main__':
# Load in list of MSD entries
msd_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'msd', 'index'))
with msd_index.searcher() as searcher:
msd_list = list(searcher.documents())
# Load in list of MSD entries
midi_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'clean_midi', 'index'))
with midi_index.searcher() as searcher:
midi_list = list(searcher.documents())
# Load in hash sequences (and metadata) for all MSD entries
msd_data = experiment_utils.load_precomputed_data(
msd_list, os.path.join(RESULTS_PATH, 'tdftm_msd_embeddings'))
# Get a list of valid MIDI-MSD match pairs
midi_msd_mapping = experiment_utils.get_valid_matches(
os.path.join(RESULTS_PATH, '{}_pairs.csv'.format(SPLIT)),
SCORE_THRESHOLD,
os.path.join(RESULTS_PATH, 'clean_midi_aligned', 'h5'))
midi_datas, midi_index_mapping = experiment_utils.load_valid_midi_datas(
midi_msd_mapping, msd_data, midi_list,
os.path.join(RESULTS_PATH, 'tdftm_clean_midi_embeddings'))
# Run match_one_midi for each MIDI data and MSD index list
results = [experiment_utils.match_embedding(
midi_datas[md5], msd_data, midi_index_mapping[md5])
for md5 in midi_datas]
# Load in network parameter values
lasagne.layers.set_all_param_values(
layers[-1], network_params['X'])
# Compile function for computing the output of the network
hash_fn = theano.function(
[layers[0].input_var],
lasagne.layers.get_output(layers[-1], deterministic=True))
# Load in list of MSD entries
msd_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'msd', 'index'))
with msd_index.searcher() as searcher:
msd_list = list(searcher.documents())
# Load MSD embeddings
msd_embedding_datas = experiment_utils.load_precomputed_data(
msd_list, os.path.join(RESULTS_PATH, 'pse_msd_embeddings'))
# Load in hash sequences (and metadata) for all MSD entries
msd_sequence_datas = experiment_utils.load_precomputed_data(
msd_list, os.path.join(RESULTS_PATH, 'dhs_msd_hash_sequences'))
# Ignore entries with very short sequences
msd_embedding_datas = [
e for d, e in zip(msd_sequence_datas, msd_embedding_datas)
if len(d['hash_sequence']) > MIN_SEQUENCE_LENGTH]
msd_sequence_datas = [d for d in msd_sequence_datas
if len(d['hash_sequence']) > MIN_SEQUENCE_LENGTH]
# Create a big matrix of the embeddings
msd_embeddings = np.concatenate(
[d['embedding'] for d in msd_embedding_datas], axis=0)
# Construct paths to each feature file
msd_feature_paths = [
os.path.join(DATA_PATH, 'msd', 'h5', d['path']) + '.h5'
if __name__ == '__main__':
# Load in list of MSD entries
msd_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'msd', 'index'))
with msd_index.searcher() as searcher:
msd_list = list(searcher.documents())
# Load in list of MSD entries
midi_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'clean_midi', 'index'))
with midi_index.searcher() as searcher:
midi_list = list(searcher.documents())
# Load in hash sequences (and metadata) for all MSD entries
msd_data = experiment_utils.load_precomputed_data(
msd_list, os.path.join(RESULTS_PATH, 'dhs_piano_msd_hash_sequences'))
# Ignore very short sequences
msd_data = [
d for d in msd_data if len(d['hash_sequence']) > MIN_SEQUENCE_LENGTH
]
# Get a list of valid MIDI-MSD match pairs
midi_msd_mapping = experiment_utils.get_valid_matches(
os.path.join(RESULTS_PATH,
'{}_pairs.csv'.format(SPLIT)), SCORE_THRESHOLD,
os.path.join(RESULTS_PATH, 'clean_midi_aligned', 'h5'))
midi_datas, midi_index_mapping = experiment_utils.load_valid_midi_datas(
midi_msd_mapping, msd_data, midi_list,
os.path.join(RESULTS_PATH, 'dhs_piano_clean_midi_hash_sequences'))
if __name__ == '__main__':
# Load in list of MSD entries
msd_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'msd', 'index'))
with msd_index.searcher() as searcher:
msd_list = list(searcher.documents())
# Load in list of MSD entries
midi_index = whoosh_search.get_whoosh_index(
os.path.join(DATA_PATH, 'clean_midi', 'index'))
with midi_index.searcher() as searcher:
midi_list = list(searcher.documents())
# Load in hash sequences (and metadata) for all MSD entries
msd_data = experiment_utils.load_precomputed_data(
msd_list, os.path.join(RESULTS_PATH, 'stats_msd_embeddings'))
# Get a list of valid MIDI-MSD match pairs
midi_msd_mapping = experiment_utils.get_valid_matches(
os.path.join(RESULTS_PATH, '{}_pairs.csv'.format(SPLIT)),
SCORE_THRESHOLD,
os.path.join(RESULTS_PATH, 'clean_midi_aligned', 'h5'))
midi_datas, midi_index_mapping = experiment_utils.load_valid_midi_datas(
midi_msd_mapping, msd_data, midi_list,
os.path.join(RESULTS_PATH, 'stats_clean_midi_embeddings'))
# Run match_one_midi for each MIDI data and MSD index list
results = [
experiment_utils.match_embedding(
midi_datas[md5], msd_data, midi_index_mapping[md5])