malar
/
speech-scoring
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

removing spec_n counter

master
Malar Kannan 2017-11-24 11:06:42 +00:00
parent ec08cc7d62
commit 43d5b75db9
3 changed files with 11 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
speech_data.py
View File

@ -210,15 +210,14 @@ def record_generator_count(records_file):
record_iterator = tf.python_io.tf_record_iterator(path=records_file) record_iterator = tf.python_io.tf_record_iterator(path=records_file)
count,spec_n = 0,0 count,spec_n = 0,0
for i in record_iterator: for i in record_iterator:
example = tf.train.Example() # example = tf.train.Example()
example.ParseFromString(i) # example.ParseFromString(i)
spec_n1 = example.features.feature['spec_n1'].int64_list.value[0] # spec_n1 = example.features.feature['spec_n1'].int64_list.value[0]
spec_n2 = example.features.feature['spec_n2'].int64_list.value[0] # spec_n2 = example.features.feature['spec_n2'].int64_list.value[0]
spec_n = max([spec_n,spec_n1,spec_n2]) # spec_n = max([spec_n,spec_n1,spec_n2])
import pdb; pdb.set_trace()
count+=1 count+=1
record_iterator = tf.python_io.tf_record_iterator(path=records_file) record_iterator = tf.python_io.tf_record_iterator(path=records_file)
return record_iterator,count,spec_n return record_iterator,count #,spec_n
def fix_csv(audio_group='audio'): def fix_csv(audio_group='audio'):
audio_csv_lines = open('./outputs/' + audio_group + '.csv','r').readlines() audio_csv_lines = open('./outputs/' + audio_group + '.csv','r').readlines()
@ -261,7 +260,8 @@ if __name__ == '__main__':
# fix_csv('story_words_test') # fix_csv('story_words_test')
#fix_csv('audio') #fix_csv('audio')
# create_spectrogram_tfrecords('story_words_test',sample_count=100,train_test_ratio=0.1) # create_spectrogram_tfrecords('story_words_test',sample_count=100,train_test_ratio=0.1)
record_generator_count() print(record_generator_count('outputs/story_phrases.full.train.tfrecords'))
print(record_generator_count('outputs/story_phrases.full.test.tfrecords'))
# create_spectrogram_tfrecords('audio',sample_count=50) # create_spectrogram_tfrecords('audio',sample_count=50)
# read_siamese_tfrecords_generator('audio') # read_siamese_tfrecords_generator('audio')
# padd_zeros_siamese_tfrecords('audio') # padd_zeros_siamese_tfrecords('audio')

4
speech_model.py
View File

@ -113,7 +113,7 @@ def train_siamese(audio_group = 'audio'):
, epochs=1000 , epochs=1000
, steps_per_epoch=epoch_n_steps , steps_per_epoch=epoch_n_steps
, validation_data=([te_pairs[:, 0], te_pairs[:, 1]], te_y) , validation_data=([te_pairs[:, 0], te_pairs[:, 1]], te_y)
, max_queue_size=32 , max_queue_size=8
, callbacks=[tb_cb, cp_cb]) , callbacks=[tb_cb, cp_cb])
model.save(model_dir+'/siamese_speech_model-final.h5') model.save(model_dir+'/siamese_speech_model-final.h5')
@ -124,4 +124,4 @@ def train_siamese(audio_group = 'audio'):
if __name__ == '__main__': if __name__ == '__main__':
train_siamese('story_phrases') train_siamese('story_phrases.full')

3
speech_test.py
View File

@ -41,7 +41,6 @@ def evaluate_siamese(records_file,audio_group='audio',weights = 'siamese_speech_
total,same_success,diff_success,skipped,same_failed,diff_failed = 0,0,0,0,0,0 total,same_success,diff_success,skipped,same_failed,diff_failed = 0,0,0,0,0,0
all_results = [] all_results = []
for (i,string_record) in tqdm(enumerate(record_iterator),total=records_count): for (i,string_record) in tqdm(enumerate(record_iterator),total=records_count):
# string_record = next(record_iterator)
total+=1 total+=1
example = tf.train.Example() example = tf.train.Example()
example.ParseFromString(string_record) example.ParseFromString(string_record)
@ -178,7 +177,7 @@ def visualize_results(audio_group='audio'):
if __name__ == '__main__': if __name__ == '__main__':
# evaluate_siamese('./outputs/story_words_test.train.tfrecords',audio_group='story_words.gpu',weights ='siamese_speech_model-58-epoch-0.00-acc.h5') # evaluate_siamese('./outputs/story_words_test.train.tfrecords',audio_group='story_words.gpu',weights ='siamese_speech_model-58-epoch-0.00-acc.h5')
# evaluate_siamese('./outputs/story_words.test.tfrecords',audio_group='story_words',weights ='siamese_speech_model-675-epoch-0.00-acc.h5') # evaluate_siamese('./outputs/story_words.test.tfrecords',audio_group='story_words',weights ='siamese_speech_model-675-epoch-0.00-acc.h5')
evaluate_siamese('./outputs/story_words_test.train.tfrecords',audio_group='story_words.gpu',weights ='siamese_speech_model-58-epoch-0.00-acc.h5') evaluate_siamese('./outputs/story_phrases.test.tfrecords',audio_group='story_phrases',weights ='siamese_speech_model-956-epoch-0.20-acc.h5')
# play_results('story_words') # play_results('story_words')
#inspect_tfrecord('./outputs/story_phrases.test.tfrecords',audio_group='story_phrases') #inspect_tfrecord('./outputs/story_phrases.test.tfrecords',audio_group='story_phrases')
# visualize_results('story_words.gpu') # visualize_results('story_words.gpu')