Heecheol Cho
Heecheol Cho
To synthesize mel spectrogram at step t(audio step), need the full (forcibly Incremental ) attention matrix. In synthesize.py, only argmaxed values are passed. Then returned _Y changes at each step....
Face Detection Data http://vis-www.cs.umass.edu/fddb/  1. Remove the 'crop part' of 'preprocess_for_train()' in ssd_vgg_preprocessing.py While cropping, all true boxes are removed, which causes problems with loss calculations. ``` dst_image, labels,...
두가지에 대해서 의견 드립니다. 1. page 386 코드중에 ``` for i in range(DEFINES.max_sequence_length): ... if params['attention'] == True: W1 = tf.keras.layers.Dense(params['hidden_size']) W2 = tf.keras.layers.Dense(params['hidden_size']) ``` attention을 계산하기 위한 Dense layer가...
How to train with multiple corpus files? Without merging files together, is it possible?
MC방법에서는 first visit, every visit 방식이 있는 것으로 알고 있습니다. 코드 구현은 last visit 방식인데, 이유가 있나요?
1-grid-word ---> 1-policy-iteration 에서 코드 전제적으로 width, height 순서가 맞지 않습니다. 코드에서는 widht=5, height=5로 되어 있어, 작동하지만, width=5, height=6이면, 작동하지 않습니다. 예들 들어, ``` self.value_table = [[0.0] * env.width for...
data_ctl.pkl, idx_phonemes.npy, idx_words.npy in cmudict-processed.tar.gz was not shuffled. Need to be shuffled.
For example, ``` self.conv1 = tf.keras.layers.Conv2D(filters=filter_num, kernel_size=(3, 3), strides=stride, padding="same") ``` shoud be ``` self.conv1 = tf.keras.layers.Conv2D(filters=filter_num, kernel_size=(3, 3), strides=stride, padding="same",use_bias=False) ```
For one iteration, two batch data are used. one batch data is used for fake_y_val, fake_x_val = sess.run([fake_y, fake_x])
Before: ``` b = scipy.misc.imresize(a[:,:,0],[84,84,1],interp='nearest') c = scipy.misc.imresize(a[:,:,1],[84,84,1],interp='nearest') d = scipy.misc.imresize(a[:,:,2],[84,84,1],interp='nearest') a = np.stack([b,c,d],axis=2) ``` After: ``` a= (skimage.transform.resize(a,[84,84,3],order=0)*255).astype(np.uint8) ```