def predict_one(items, outs, model_names, config):
data, bpm_label, beat_label, downbeat_label, song_name = items
import matplotlib.pyplot as plt
# plt.imshow(data[0].T, aspect=20, origin="lower")
# plt.show()
# plt.savefig("/home/ooyama/melspectrogram.pdf")
# exit()
bpmgram, beat, downbeat = outs
# plt.plot(softmax(bpmgram))
# plt.show()
# plot_tempogram(bpmgram)
bpm = np.argmax(softmax(bpmgram), axis=-1)
print(bpm.shape, beat.shape)
plot_beat_and_bpm2(
{
"beat": (1, beat),
"beat_label": (2, beat_label),
"downbeat": (3, downbeat),
# "downbeatlabel": (4, downbeat_label),
},
song_name,
model_names,
time_range=(2500, 3500),
bpm_range=(50, 200),
)
def predict_one(items, outs, model_names, config):
data, beattheta, bpm_label, song_name = items
theta, idxs = numpy_fft(outs[None, :] * 2 * np.pi, 100)
plot_beat_and_bpm2(
# theta[0],
# {"beat": outs, "beat_label": beattheta},
{
"before": outs,
"after": theta[0]
},
song_name=song_name,
model_names=model_names,
time_range=(0, 6000),
)
k = 200
sin = np.sin(outs * 2 * np.pi)
spectrogram = [
np.abs(np.fft.fft(sin[i:i + k]))[:k // 2]
for i in range(outs.shape[0] - k)
]
spectrogram = np.stack(spectrogram)
# plt.plot(np.abs(np.fft.fft(sin))[:3000])
print(np.argmax(spectrogram[4]))
plt.imshow(spectrogram.T, aspect=40, origin="lower")
# plt.plot(spectrogram[5])
plt.show()
def predict_one(items, outs, model_names, config):
data, beattheta, downbeat_label, song_name = items
# theta, idxs = numpy_fft(outs[None, :] * 2 * np.pi, 100)
plot_beat_and_bpm2(
# beat=outs,
# beat_label=downbeat_label,
# theta[0],
{"downbeat": outs, "downbeat_label": downbeat_label, "beat_label":beattheta},
# {"before": outs, "after": theta[0]},
song_name=song_name,
model_names=model_names,
time_range=(0, 6000),
)
def predict_one(items, outs, model_names, config):
data, bpm_label, beat_label, song_name = items
bpmgram, beat = outs
bpmgram = bpmgram.T
# plot_tempogram(bpmgram)
bpm = np.argmax(bpmgram, axis=-1)
print(bpm.shape, beat.shape)
plot_beat_and_bpm2(
{"beat": (1, beat)},
song_name,
model_names,
time_range=(0000, 1000),
bpm_range=(50, 200),
)
def predict_one(items, outs, model_names, config):
data, beattheta, downbeat_label, song_name = items
beat, downbeat = outs
plt.plot(beat)
plt.plot(downbeat)
plt.show()
theta, idxs = numpy_fft(beat[None, :] * 2 * np.pi, 400)
plot_beat_and_bpm2(
# theta[0],
{"beat": beat, "downbeat_label": downbeat_label, "downbeat": downbeat},
# {"downbeat": theta[0]/2/np.pi, "label": beattheta},
song_name=song_name,
model_names=model_names,
time_range=(0, 6000),
)
def predict_one(items, outs, model_names, config):
data, beattheta, bpm_label, song_name = items
theta, f_logsoftmax = outs
f_logsoftmax = torch.from_numpy(f_logsoftmax)
bpm_label = torch.from_numpy(bpm_label)
# theta, idxs = numpy_fft(outs[None, :] * 2 * np.pi, 100)
f_label = get_bpm_label(bpm_label[None, :], f_logsoftmax[None, :])[0]
# exit()
plot_beat_and_bpm2(
# theta[0],
{"beat": theta, "beat_label": beattheta},
# {"beat": np.sin(theta*2*np.pi)},
# {"a": f_label[2000], "before": torch.exp(f_logsoftmax[2000]),},
song_name=song_name,
model_names=model_names,
# time_range=(0, 6000),
)
def predict_one(items, outs, model_names, config):
data, bpm_label, beattheta_label, downbeattheta_label, song_name = items
beattheta_out, downbeattheta_out, deltabeat_out, deltadownbeat_out = outs
diffbeattheta = medfilt(deltabeat_out, kernel_size=21)
beattheta_out = np.argmax(beattheta_out.T, axis=-1) / 6000
downbeattheta_out = np.argmax(downbeattheta_out.T, axis=-1) / 6000
activation = beattheta_out[:-1] - beattheta_out[1:]
activation[activation <= 0] = 1e-6
bpm_out = diffbeattheta * 3000 / np.pi
plot_beat_and_bpm2(
{
"beat": (1, beattheta_out),
"beat_label": (2, beattheta_label),
"downbeat": (3, downbeattheta_out),
# "bpm": (4, bpm_out),
},
song_name,
model_names,
time_range=(2500, 3500),
bpm_range=(50, 200),
)