# Prepare Theano variables for inputs and targets input_var = T.tensor4('inputs') # restore network network = build_mlp(input_var) network_parameters = np.load(FILE) lasagne.layers.set_all_param_values(network, network_parameters['arr_0']) # prepare prediction test_prediction = lasagne.layers.get_output(network, deterministic=True) predict_fn = theano.function([input_var], T.argmax(test_prediction, axis=1)) complete_x = list() complete_y = list() for position, fft, c in traindata_mix.test_data_iterator( traindata_mix.RING_02_TEST_DATA): try: complete_x.append(predict_fn([[[fft]]]) * 10000) except: print position, fft complete_x.append(0) complete_x.append(0) rate, data = read(traindata_mix.RING_02_TEST_DATA + ".wav") plot(range(0, len(data)), data) plot(range(0, len(complete_x)), complete_x) show()
dtype=theano.config.floatX) # prepare prediction test_prediction = lasagne.layers.get_output(network, deterministic=True) predict_fn = theano.function([input_var], T.argmax(test_prediction, axis=1)) # Compile a second function computing the validation loss and accuracy: val_fn = theano.function([input_var, target_var], [test_loss, test_acc], allow_input_downcast=True) complete_x = list() complete_y = list() a = 0 correct = 0 wrong = 0 for position, fft, c in traindata_mix.test_data_iterator(traindata_mix.RING_01_TEST_DATA): try: r = predict_fn([[[fft]]]) if c > 0: a += 1 if r > 0 and c > 0: correct += 1 if c > 0 >= r: wrong += 1 complete_x.append(r * 10000) except: print position, fft complete_x.append(0) complete_x.append(0)