experiments/STT
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[FIX] reversible error if dest_time == 0

Browse Source
This commit is contained in:
Yi-Hua Chiu 2019-12-18 18:22:38 +08:00
parent 533d15645f
commit 72c09ebb38
1 changed files with 4 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
util/spectrogram_augmentations.py
View File

@ -105,14 +105,10 @@ def augment_sparse_warp(spectrogram, time_warping_para=80, interpolation_order=2
source_min = tf.math.minimum(source_max - 1, time_warping_para)
rand_source_time = tfv1.random_uniform( # generate source points `t` of time axis between (W, tau-W)
[], source_min, source_max, tf.int32)
rand_dest_time = tfv1.random_uniform( # generate dest points `t'` of time axis between (t-W, t+W)
[], tf.math.maximum(tf.math.subtract(rand_source_time, time_warping_para), 0), tf.math.add(rand_source_time, time_warping_para), tf.int32)
rand_dest_time = tfv1.random_uniform( # generate dest points `t'` of time axis between (t-W, t+W), !!! if rand_dest_time == 0, might raise invertible error
[], tf.math.maximum(tf.math.subtract(rand_source_time, time_warping_para), 1), tf.math.add(rand_source_time, time_warping_para), tf.int32)
# if choosen_freq == tau -1 => crash
choosen_freq = tf.cond(tf.equal(choosen_freqs[i], tau-1),
lambda: choosen_freqs[i] + # pylint: disable=cell-var-from-loop
1, # pylint: disable=cell-var-from-loop
lambda: choosen_freqs[i]) # pylint: disable=cell-var-from-loop
choosen_freq = choosen_freqs[i]
sources.append([0, choosen_freq])
sources.append([rand_source_time, choosen_freq])
sources.append([tau, choosen_freq])
@ -126,7 +122,7 @@ def augment_sparse_warp(spectrogram, time_warping_para=80, interpolation_order=2
dest_control_point_locations = tf.cast([dests], tf.float32)
# debug
# print('spectrogram', spectrogram)
# spectrogram = tf.Print(spectrogram, [tf.shape(spectrogram)], message='spectrogram', first_n=1000)
# spectrogram = tf.Print(spectrogram, sources, message='sources', first_n=1000)
# spectrogram = tf.Print(spectrogram, dests, message='dests', first_n=1000)