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

updated model data

master
Malar Kannan 2017-11-27 14:08:01 +05:30
parent 43d5b75db9
commit 20b2d7a958
4 changed files with 8 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.gitignore vendored
View File

@ -143,3 +143,5 @@ inputs/audio*
logs/* logs/*
models/* models/*
*.pkl *.pkl
temp/*
trained/*

7
speech_data.py
View File

@ -147,7 +147,7 @@ def read_siamese_tfrecords_generator(audio_group='audio',batch_size=32,test_size
(n_spec,n_features,n_records) = pickle.load(open(const_file,'rb')) (n_spec,n_features,n_records) = pickle.load(open(const_file,'rb'))
def copy_read_consts(dest_dir): def copy_read_consts(dest_dir):
shutil.copy2(const_file,dest_dir) shutil.copy2(const_file,dest_dir+'/constants.pkl')
return (n_spec,n_features,n_records) return (n_spec,n_features,n_records)
# @threadsafe_iter # @threadsafe_iter
def record_generator(): def record_generator():
@ -259,9 +259,8 @@ if __name__ == '__main__':
# create_spectrogram_tfrecords('story_all',sample_count=25) # create_spectrogram_tfrecords('story_all',sample_count=25)
# 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=10,train_test_ratio=0.1)
print(record_generator_count('outputs/story_phrases.full.train.tfrecords')) #record_generator_count()
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')

2
speech_model.py
View File

@ -124,4 +124,4 @@ def train_siamese(audio_group = 'audio'):
if __name__ == '__main__': if __name__ == '__main__':
train_siamese('story_phrases.full') train_siamese('story_words_test')

4
speech_test.py
View File

@ -29,7 +29,7 @@ def test_with(audio_group):
def evaluate_siamese(records_file,audio_group='audio',weights = 'siamese_speech_model-final.h5'): def evaluate_siamese(records_file,audio_group='audio',weights = 'siamese_speech_model-final.h5'):
# audio_group='audio';model_file = 'siamese_speech_model-305-epoch-0.20-acc.h5' # audio_group='audio';model_file = 'siamese_speech_model-305-epoch-0.20-acc.h5'
# records_file = os.path.join('./outputs',eval_group+'.train.tfrecords') # records_file = os.path.join('./outputs',eval_group+'.train.tfrecords')
const_file = os.path.join('./models/'+audio_group+'/',audio_group+'.constants') const_file = os.path.join('./models/'+audio_group+'/','constants.pkl')
arch_file='./models/'+audio_group+'/siamese_speech_model_arch.yaml' arch_file='./models/'+audio_group+'/siamese_speech_model_arch.yaml'
weight_file='./models/'+audio_group+'/'+weights weight_file='./models/'+audio_group+'/'+weights
(n_spec,n_features,n_records) = pickle.load(open(const_file,'rb')) (n_spec,n_features,n_records) = pickle.load(open(const_file,'rb'))
@ -177,7 +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_phrases.test.tfrecords',audio_group='story_phrases',weights ='siamese_speech_model-956-epoch-0.20-acc.h5') evaluate_siamese('./outputs/story_words_test.train.tfrecords',audio_group='story_words_test.10',weights ='siamese_speech_model-891-epoch-0.02-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')