def evaluate(items, outs, config, trial=None):
eval_bpm = EvaluateBpm()
eval_beat = EvaluateBeat("beat")
data, bpm_label, beat_label = items
downbeattheta_out, delta_beattheta = outs
# beat estimation
downbeattheta_out = np.argmax(downbeattheta_out, axis=1) / 6000 * 2 * math.pi
beat_metrics = eval_beat.batch(downbeattheta_out, beat_label)
# # global bpm estimation
# bpm_out = delta_beattheta * 3000 / math.pi
# bpm_out = medfilt(bpm_out, kernel_size=[1, 21])
# m, cnt = mode(np.round(bpm_out).astype(np.int64), axis=1)
bpm_label = np.mean(bpm_label, axis=-1)
m = calc_bpm_from_theta(downbeattheta_out, "beat")
bpm_metrics = eval_bpm.global_(m, bpm_label)
# local bpm estimation
# bpm_metrics = eval_bpm.local(bpm_out, bpm_label)
ret = {"beat": beat_metrics, "bpm": bpm_metrics}
return ret
return ret
def evaluate(items, outs, config):
eval_bpm = EvaluateBpm()
eval_beat = EvaluateBeat(estimated_type="activation",
target_type="activation")
data, bpm_label, beat_label = items
beattheta_out, delta_beattheta = outs
# beat estimation
beattheta_out = np.argmax(beattheta_out, axis=1) / 6000
# theta = numpy_fft(beattheta_out * 2 * np.pi, 500)
# theta /= (2 * np.pi)
# activation = theta[:, :-1] - theta[:, 1:]
# activation[activation <= 0] = 1e-6
beat_metrics = eval_beat.batch(beattheta_out, beat_label)
# beat_metrics = eval_beat.batch(activation, beat_label)
# global bpm estimation
bpm_out = delta_beattheta * 3000 / np.pi
bpm_out = medfilt(bpm_out, kernel_size=[1, 21])
m = calc_bpm_from_theta(beattheta_out * 2 * np.pi, "beat")
bpm_label = np.mean(bpm_label, axis=-1)
bpm_metrics = eval_bpm.global_(m, bpm_label)
ret = {"beat": beat_metrics, "bpm": bpm_metrics}
return ret
def evaluate(items, outs, config, data_type="test"):
eval_bpm = EvaluateBpm()
eval_beat = EvaluateBeat(estimated_type="activation",
target_type="activation")
eval_downbeat = EvaluateDownbeat(estimated_type="activation",
target_type="activation")
data, bpm_label, beattheta_label, downbeattheta_label = items
beattheta_out, downbeattheta_out, deltabeat_out, deltadownbeat_out = outs
# beat estimation
beattheta_out = np.argmax(beattheta_out, axis=1) / 6000
downbeattheta_out = np.argmax(downbeattheta_out, axis=1) / 6000
activation = beattheta_out[:, :-1] - beattheta_out[:, 1:]
activation[activation <= 0.7] = 1e-6
beat_metrics = eval_beat.batch(beattheta_out, beattheta_label)
downbeat_met = eval_downbeat.batch(downbeattheta_out, downbeattheta_label)
# global bpm estimation
bpm_out = deltabeat_out * 3000 / np.pi
bpm_out = medfilt(bpm_out, kernel_size=[1, 21])
if data_type == "valid":
m, cnt = mode(np.round(bpm_out).astype(np.int64), axis=1)
m = m[:, 0]
elif data_type == "test":
m = calc_bpm_from_theta(beattheta_out * 2 * np.pi, "beat")
bpm_label = np.mean(bpm_label, axis=-1)
bpm_metrics = eval_bpm.global_(m, bpm_label)
ret = {"beat": beat_metrics, "bpm": bpm_metrics, "downbeat": downbeat_met}
return ret
def evaluate(items, outs, config, trial=None):
data, beat_label = items
beattheta_out = outs
eval_beat = EvaluateBeat(estimated_type="activation",
target_type="activation")
beattheta_out = np.argmax(beattheta_out, axis=1) / 6000
beat_metrics = eval_beat.batch(beattheta_out, beat_label)
ret = {"beat": beat_metrics}
return ret
def evaluate(items, outs, config, trial=None):
data, bpm_label, beat_label = items
beat_out, bpm_out = outs
eval_beat = EvaluateBeat("downbeat")
eval_bpm = EvaluateBpm()
beat_metrics = eval_beat.batch(beat_out, beat_label)
bpm_metrics = eval_bpm.local(bpm_out, bpm_label)
ret = {"beat": beat_metrics, "bpm": bpm_metrics}
return ret
def evaluate(items, outs, config, trial=None):
data, bpm_label, beat_label = items
bpm_out, beat_out = outs # (b,300,T), (b,T)
bpm_out = np.mean(bpm_out, axis=-1)
bpm_out = np.argmax(bpm_out, axis=1)
bpm_label = np.mean(bpm_label, axis=-1)
eval_beat = EvaluateBeat("beat")
eval_bpm = EvaluateBpm()
beat_metrics = eval_beat.batch(beat_out, beat_label)
bpm_metrics = eval_bpm.global_(bpm_out, bpm_label)
ret = {"beat": beat_metrics, "bpm": bpm_metrics}
return ret
def evaluate(items, outs, config, trial=None):
data, bpm_label, beat_label = items
bpm_out, beat_out = outs
bpm_out = np.argmax(bpm_out, axis=1)
eval_beat = EvaluateBeat(estimated_type="activation",
target_type="activation")
eval_bpm = EvaluateBpm()
beat_metrics = eval_beat.batch(beat_out, beat_label)
bpm_label = np.mean(bpm_label, axis=-1)
bpm_metrics = eval_bpm.global_(bpm_out, bpm_label)
ret = {"beat": beat_metrics, "bpm": bpm_metrics}
return ret
def evaluate(items, outs, config, trial=None):
eval_beat = EvaluateBeat(estimated_type="activation",
target_type="activation")
data, beat_label = items
sintheta, costheta = outs
theta = np.arctan2(sintheta, costheta)
theta[theta < 0] += 2 * np.pi
beat_metrics = eval_beat.batch(theta, beat_label)
ret = {"beat": beat_metrics}
return ret
def evaluate(items, outs, config, trial=None):
# eval_beat = EvaluateBeat(estimated_type="index", target_type="activation")
eval_beat = EvaluateBeat(estimated_type="activation", target_type="activation")
data, beat_label, bpm_label = items
beattheta, f_softmax = outs
# theta, idxs = numpy_fft(outs * 2 * np.pi, 100)
# act = (outs[:, :-1] - outs[:, 1:]) / (2 * np.pi)
# act[act <= 0] = 1e-5
beat_metrics = eval_beat.batch(beattheta, beat_label)
ret = {"beat": beat_metrics}
return ret
def evaluate(items, outs, config, trial=None):
eval_bpm = EvaluateBpm()
eval_beat = EvaluateBeat(estimated_type="activation", target_type="activation")
data, bpm_label, beat_label = items
beattheta_out, delta_beattheta = outs
# beattheta_out, idxs = numpy_fft(beattheta_out * 2 * np.pi, 100)
# beat estimation
beat_metrics = eval_beat.batch(beattheta_out, beat_label)
# global bpm estimation
bpm_out = delta_beattheta * 3000 / np.pi
bpm_out = medfilt(bpm_out, kernel_size=[1, 21])
m = calc_bpm_from_theta(beattheta_out * 2 * np.pi, "beat")
bpm_label = np.mean(bpm_label, axis=-1)
bpm_metrics = eval_bpm.global_(m, bpm_label)
ret = {"beat": beat_metrics, "bpm": bpm_metrics}
return ret
def evaluate(items, outs, config, trial=None):
eval_bpm = EvaluateBpm()
eval_beat = EvaluateBeat("beat")
data, bpm_label, beat_label = items
beat_out, delta_beattheta = outs
bpm_out = delta_beattheta * 3000 / math.pi
bpm_out = medfilt(bpm_out, kernel_size=[1, 21])
beat_metrics = eval_beat.batch(beat_out, beat_label)
# global bpm estimation
m, cnt = mode(np.round(bpm_out).astype(np.int64), axis=1)
bpm_label = np.mean(bpm_label, axis=-1)
bpm_metrics = eval_bpm.global_(m, bpm_label)
# local bpm estimation
# bpm_metrics = eval_bpm.local(bpm_out, bpm_label)
ret = {"beat": beat_metrics, "bpm": bpm_metrics}
return ret
def evaluate(items, outs, config, trial=None):
eval_bpm = EvaluateBpm()
eval_beat = EvaluateBeat(estimated_type="activation",
target_type="activation")
eval_downbeat = EvaluateDownbeat(estimated_type="activation",
target_type="activation")
data, bpm_label, beat_label, downbeat_label = items
bpm_out, beat_out, downbeat_out = outs
bpm_out = np.argmax(bpm_out, axis=1)
beat_out = beat_out - downbeat_out
beat_out[beat_out <= 0] = 1e-6
beat_metrics = eval_beat.batch(beat_out, beat_label)
downbeat_metrics = eval_downbeat.batch(downbeat_out, downbeat_label)
bpm_label = np.mean(bpm_label, axis=-1) # (batchsize,T) -> (batchsize)
bpm_metrics = eval_bpm.global_(bpm_out, bpm_label)
ret = {
"beat": beat_metrics,
"downbeat": downbeat_metrics,
"bpm": bpm_metrics
}
return ret
