def cpu(*F, N=(500, 5000, 500)):
'''Mostra gráfico com o consumo de tempo das funções *F, para
entradas com tamanho variando no intervalo N=(500,5000,500).
Um parâmetro E = g define a função que gera as entradas'''
C = ['r', 'b', 'g', 'c', 'm', 'y', 'k', 'w']
M = ['o', 's', 'p', 'h', '+', 'x', 'v', '^']
X = []
Y = [[] for _ in range(len(F))]
print('n = ', end='')
inicio = time()
rodada = 1
for n in range(N[0], N[1] + 1, N[2]):
print('%s ' % rodada, end='')
X.append(n)
v = sample(range(0, n), n)
for i, f in enumerate(F):
w = deepcopy(v)
Y[i].append(tempo(f, w))
rodada += 1
print('\nTempo total: %.1fs' % (time() - inicio))
xlabel('tamanho da entrada')
ylabel('tempo de execução (s)')
ticklabel_format(style='sci', axis='both', scilimits=(-3, +4))
grid(True)
for i, f in enumerate(F):
plot(X,
Y[i],
color=C[i % len(C)],
marker=M[i % len(M)],
label=f.__name__)
legend(loc='upper left')
show()
# Read the labeled tempo
bpm = float(utils.read_tempofile(DB, f))
print('ground-truth tempo: ', bpm)
label.append(bpm)
# Estimate a static tempo
sr, y = utils.read_wav(f)
hop_length = 512
oenv = librosa.onset.onset_strength(y=y, sr=sr, hop_length=hop_length)
# tempogram = librosa.feature.tempogram(onset_envelope=oenv, sr=sr, hop_length=hop_length)
# predict the tempo1(slower one), tempo2(faster one)
# tempo1, tempo2 = librosa.beat.tempo(onset_envelope=oenv, sr=sr, hop_length=hop_length)
tempo1, tempo2 = utils.tempo(onset_envelope=oenv, sr=sr, hop_length=hop_length)
pred_t1.append(tempo1)
pred_t2.append(tempo2)
print(tempo1, tempo2)
# p score
s1 = tempo1/(tempo1+tempo2)
s2 = 1.0 - s1
print(s1, s2)
p = s1 * utils.P_score(tempo1, bpm) + s2 * utils.P_score(tempo2, bpm)
p_score.append(p)
# ALOTC score
ALOTC = utils.ALOTC(tempo1, tempo2, bpm)
ALOTC_score.append(ALOTC)
def executar_teste(t): #t=(função, lista)
print('processando %s com uma lista de %d' % (t[0].__name__, len(t[1])))
return tempo(t[0], t[1])
def beat_estimate(DB,GENRE,fout,name="***** Q1 *****\n",multi=1,method="own"):
genres_p, genres_ALOTC = list(), list()
for genre in GENRE:
print('GENRE:', genre)
FILES = glob(DB + '/wav/' + genre + '/*.wav')
label, p_score, ALOTC_score = list(), list(), list()
for f in FILES:
f = f.replace('\\', '/')
#print('FILE:', f)
# Read the labeled tempo
bpm = float(utils.read_tempofile(DB, genre, f))
#print('ground-truth tempo: ', bpm)
label.append(bpm)
if method=="own":
# Estimate a static tempo
sr, y = utils.read_wav(f)
hop_length = 512
oenv = librosa.onset.onset_strength(y=y, sr=sr, hop_length=hop_length)
# predict the tempo1(slower one), tempo2(faster one)
tempo1, tempo2 = utils.tempo(onset_envelope=oenv, sr=sr, hop_length=hop_length)
elif method=="madmom":
# madmom estimate tempo
proc = madmom.features.tempo.TempoEstimationProcessor(fps=100)
act = madmom.features.beats.RNNBeatProcessor()(f)
tempo1 = (proc(act)).astype(float)[0][0].item()
tempo2 = (proc(act)).astype(float)[1][0].item()
# Q2
tempo1 = tempo1*multi
tempo2 = tempo2*multi
# p score
s1 = tempo1/(tempo1+tempo2)
s2 = 1.0 - s1
#print(s1, s2)
p = s1 * utils.P_score(tempo1, bpm) + s2 * utils.P_score(tempo2, bpm)
p_score.append(p)
# ALOTC score
ALOTC = utils.ALOTC(tempo1, tempo2, bpm)
ALOTC_score.append(ALOTC)
#print(p, ALOTC)
p_avg = sum(p_score)/len(p_score)
ALOTC_avg = sum(ALOTC_score)/len(ALOTC_score)
genres_p.append(p_avg)
genres_ALOTC.append(ALOTC_avg)
print('finished : ',genre)
print(genres_p)
print(genres_ALOTC)
fout.write(name)
fout.write("Genre \tP-score \tALOTC score\n")
for genre in range(len(GENRE)):
fout.write("{:13s}\t{:8.2f}\t{:8.2f}\n".format(GENRE[genre], genres_p[genre], genres_ALOTC[genre]))
fout.write('----------\n')
fout.write("Overall P-score:\t{:.2f}\n".format(sum(genres_p)/len(genres_p)))
fout.write("Overall ALOTC score:\t{:.2f}\n".format(sum(genres_ALOTC)/len(genres_ALOTC)))