search

yltw27 / KeywordSpotting

Train a 4-layer Convolutional Neural Network to detect trigger word
52 stars 18 forks source link

No change in accuracy even with 200 samples #2

Closed cogmeta closed 6 years ago

cogmeta commented 6 years ago

('Epoch:', 10) Training Accuracy = 48.438 % Training Loss = 0.693659 Validation Accuracy = 43.750 % Validation Loss = 0.696315 ('Checkpoint:', 'ckpt/2018-04-21/kws_v1-20180421-1025-10')

('Epoch:', 11) Training Accuracy = 48.438 % Training Loss = 0.693522 Validation Accuracy = 43.750 % Validation Loss = 0.696481 ('Checkpoint:', 'ckpt/2018-04-21/kws_v1-20180421-1025-11')

('Epoch:', 12) Training Accuracy = 48.438 % Training Loss = 0.693392 Validation Accuracy = 43.750 % Validation Loss = 0.696657 ('Checkpoint:', 'ckpt/2018-04-21/kws_v1-20180421-1025-12')

('Epoch:', 13) Training Accuracy = 48.438 % Training Loss = 0.693266 Validation Accuracy = 43.750 % Validation Loss = 0.696841 ('Checkpoint:', 'ckpt/2018-04-21/kws_v1-20180421-1026-13')

('Epoch:', 14) Training Accuracy = 48.438 % Training Loss = 0.693142 Validation Accuracy = 43.750 % Validation Loss = 0.697036 ('Checkpoint:', 'ckpt/2018-04-21/kws_v1-20180421-1026-14')