def get_tempo_info(file_number):
# get data
d = BallroomData()
# wav name
wav_name = d.file_name(file_number)
# ground truth
aaagt_bpm = d.get_gt_bpm(file_number)
# load file for librosa
y, sr = d[file_number]
onset_env = librosa.onset.onset_strength(y, sr=sr)
# Static tempo
aaatempo = librosa.beat.tempo(onset_envelope=onset_env, sr=sr)
# Dynamic tempo
dtempo = librosa.beat.tempo(onset_envelope=onset_env, sr=sr, aggregate=None)
# Dynamic tempo distribution
zapple = get_tempo_distribu(dtempo)
aaatempo_list = zapple[0]
aaatempo_cnt = zapple[1]
# # --- print info ---
# print("filename = ", wav_name)
# print("gt = ", aaagt_bpm)
# print("static = ", aaatempo[0])
# print("----- guess ------ ------")
# dis_lan = len(aaatempo_cnt)
# for i in range(dis_lan):
# print("%18.14f" % aaatempo_list[i], "%6d" % aaatempo_cnt[i])
# # print(str(aaatempo_list[i]), str(aaatempo_cnt[i]))
# print("----- ----- ------ ------\n\n")
# # --- ----- ---- ---
return aaagt_bpm, aaatempo, aaatempo_list, aaatempo_cnt
def get_tempo_info(file_number):
# get data
d = BallroomData()
# wav name
wav_name = d.file_name(file_number)
# ground truth
aaagt_bpm = d.get_gt_bpm(file_number)
# load file for librosa
y, sr = d[file_number]
onset_env = librosa.onset.onset_strength(y, sr=sr)
# Static tempo
aaatempo = librosa.beat.tempo(onset_envelope=onset_env, sr=sr)
# Dynamic tempo
dtempo = librosa.beat.tempo(onset_envelope=onset_env, sr=sr, aggregate=None)
# Dynamic tempo distribution
zapple = get_tempo_distribu(dtempo)
aaatempo_list = zapple[0]
aaatempo_cnt = zapple[1]
return aaagt_bpm, aaatempo, aaatempo_list, aaatempo_cnt
s = D.shape[1] # 2048 22050
f_tempo_list = []
for t in range(D.shape[1]):
# get every most tempo and list them
f_tempo_list.append(
ff_get_two_tempo(D, t, total_time)[0][0])
# get t1 t2 by using the most exist
tempo_list, tempo_cnt = get_tempo_distribu(
f_tempo_list)
t_1, t_2 = f_get_two_tempo_total(tempo_list, tempo_cnt)
# TODO: remember gt
gt = d.get_gt_bpm(file_number)
print("gt =", gt, "t1 =", t_1, " t2 =", t_2)
# calculat saliency
ps = []
for ttt in range(D.shape[1]):
# fet ft
ft1, ft2 = get_two_f_t(D, ttt, total_time, t_1,
t_2)
# get p-score
ps.append(p_score_d4(t_1, t_2, ft1, ft2, gt))
# # print ps list
# print("---p-score list---\n", ps)
R_total = []
F_total = []
for song_id in range(len(dataset)):
genre = dataset.get_genre(song_id)
if genre != "Waltz":
print(genre)
continue
aud, sr = dataset[0]
bpm = librosa.beat.tempo(aud)
beats_gt, beats_num = dataset.get_gg(song_id)
print(bpm, dataset.get_gt_bpm(song_id))
onset_frames = librosa.onset.onset_detect(y=aud, sr=sr)
librosa.frames_to_time(onset_frames, sr=sr)
o_env = librosa.onset.onset_strength(aud, sr=sr)
# o_env = scipy.signal.medfilt(o_env,3)
o_env = sigmoid(o_env * 0.5)
tempo, beats = librosa.beat.beat_track(onset_envelope=o_env, sr=sr)
times = librosa.frames_to_time(np.arange(len(o_env)), sr=sr)
onset_frames = librosa.onset.onset_detect(onset_envelope=o_env, sr=sr)
# plt.figure()
#
# plt.title('Power spectrogram')
if __name__ == '__main__':
total_of_total = 0
d = BallroomData()
for target_tempo in range(119, 123, 1):
t_zapple = [], []
total = 0
print("target tempo = ", target_tempo)
for file_number in range(len(d) - 1):
# ground truth
aaagt_bpm = d.get_gt_bpm(file_number)
if aaagt_bpm == target_tempo:
# wav name
wav_name = d.file_name(file_number)
# ground truth
aaagt_bpm = d.get_gt_bpm(file_number)
# load file for librosa
y, sr = d[file_number]
onset_env = librosa.onset.onset_strength(y, sr=sr)
# Static tempo
aaatempo = librosa.beat.tempo(onset_envelope=onset_env, sr=sr)
# Dynamic tempo
dtempo = librosa.beat.tempo(onset_envelope=onset_env,
sr=sr,
# 50
aud, sr = d[song_id]
total_time = aud.shape[0] / sr
D = fourier_tempogram(aud)
s = D.shape[1] # 2048 20050
tempo_list = []
for t in range(D.shape[1]):
tempo_list.append(f_get_two_tempo(D, t, total_time)[0][0])
# print("f_get_two_tempo", end=' : ')
# print("- - - - - - - - - - - - - - - - - - - - - - - - - - "
# "(t%d)" % t, f_get_two_tempo(D, t, total_time))
print(".")
tempo_list, tempo_cnt = get_tempo_distribu(tempo_list)
t_1, t_2 = f_get_two_tempo_tatal(tempo_list, tempo_cnt)
print("t1 = ", t_1, " t2 = ", t_2)
# print(dataset.get_genre(song_id))
print("gt = ", end='')
print(d.get_gt_bpm(song_id))
librosa.display.specshow(D[20:, :])
# plt.show()
print(".")
# dataset = StdData()
P_total = []
R_total = []
F_total = []
genres = ["Cha Cha", "Slow Waltz"]
for genre in genres:
rng_h, rng_t = qui_genre_rng(genre)
print(genre)
for song_id in range(rng_h, rng_t):
aud, sr = dataset[0]
bpm = librosa.beat.tempo(aud)
beats_gt, beats_num = dataset.get_gg(song_id)
print("bpm =", bpm, "gt =", dataset.get_gt_bpm(song_id), end=',')
onset_frames = librosa.onset.onset_detect(y=aud, sr=sr)
librosa.frames_to_time(onset_frames, sr=sr)
o_env = librosa.onset.onset_strength(aud, sr=sr)
# o_env = scipy.signal.medfilt(o_env,3)
o_env = sigmoid(o_env * 0.5)
tempo, beats = librosa.beat.beat_track(onset_envelope=o_env, sr=sr)
times = librosa.frames_to_time(np.arange(len(o_env)), sr=sr)
onset_frames = librosa.onset.onset_detect(onset_envelope=o_env,
sr=sr)
beats = times[beats]