From 43d5b75db92a3b62188ad6d4b944d139729df228 Mon Sep 17 00:00:00 2001 From: Malar Kannan Date: Fri, 24 Nov 2017 11:06:42 +0000 Subject: [PATCH] removing spec_n counter --- speech_data.py | 16 ++++++++-------- speech_model.py | 4 ++-- speech_test.py | 3 +-- 3 files changed, 11 insertions(+), 12 deletions(-) diff --git a/speech_data.py b/speech_data.py index 4df9aa2..9b3ea54 100644 --- a/speech_data.py +++ b/speech_data.py @@ -210,15 +210,14 @@ def record_generator_count(records_file): record_iterator = tf.python_io.tf_record_iterator(path=records_file) count,spec_n = 0,0 for i in record_iterator: - example = tf.train.Example() - example.ParseFromString(i) - spec_n1 = example.features.feature['spec_n1'].int64_list.value[0] - spec_n2 = example.features.feature['spec_n2'].int64_list.value[0] - spec_n = max([spec_n,spec_n1,spec_n2]) - import pdb; pdb.set_trace() + # example = tf.train.Example() + # example.ParseFromString(i) + # spec_n1 = example.features.feature['spec_n1'].int64_list.value[0] + # spec_n2 = example.features.feature['spec_n2'].int64_list.value[0] + # spec_n = max([spec_n,spec_n1,spec_n2]) count+=1 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'): audio_csv_lines = open('./outputs/' + audio_group + '.csv','r').readlines() @@ -261,7 +260,8 @@ if __name__ == '__main__': # fix_csv('story_words_test') #fix_csv('audio') # 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) # read_siamese_tfrecords_generator('audio') # padd_zeros_siamese_tfrecords('audio') diff --git a/speech_model.py b/speech_model.py index f083360..ff2822e 100644 --- a/speech_model.py +++ b/speech_model.py @@ -113,7 +113,7 @@ def train_siamese(audio_group = 'audio'): , epochs=1000 , steps_per_epoch=epoch_n_steps , validation_data=([te_pairs[:, 0], te_pairs[:, 1]], te_y) - , max_queue_size=32 + , max_queue_size=8 , callbacks=[tb_cb, cp_cb]) model.save(model_dir+'/siamese_speech_model-final.h5') @@ -124,4 +124,4 @@ def train_siamese(audio_group = 'audio'): if __name__ == '__main__': - train_siamese('story_phrases') + train_siamese('story_phrases.full') diff --git a/speech_test.py b/speech_test.py index ddf2d7b..26f5a54 100644 --- a/speech_test.py +++ b/speech_test.py @@ -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 all_results = [] for (i,string_record) in tqdm(enumerate(record_iterator),total=records_count): - # string_record = next(record_iterator) total+=1 example = tf.train.Example() example.ParseFromString(string_record) @@ -178,7 +177,7 @@ def visualize_results(audio_group='audio'): 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.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') #inspect_tfrecord('./outputs/story_phrases.test.tfrecords',audio_group='story_phrases') # visualize_results('story_words.gpu')