conv_out3=MyConvnetLayer(W3,B3,input=conv_out2.output,filter_shape=(128, 256, 3, 3),image_shape=(batch_size, 256, 10, 10),conv_stride=(1,1))
conv_out4=MyConvnetLayer(W4,B4,input=conv_out3.output,filter_shape=(128, 128, 3, 3),image_shape=(batch_size, 128, 8, 8),conv_stride=(1,1))
layer5_input = conv_out4.output.flatten(2)
# construct a fully-connected sigmoidal layer
full_5 = HiddenLayer(
W5,B5,
input=layer5_input,
n_in=128 * 6 * 6,
n_out=256,
activation=T.tanh
)
# classify the values of the fully-connected sigmoidal layer
full_5_softmax = LogisticRegression(W6,B6,input=full_5.output, n_in=256, n_out=5)
weight_decay=1e-5
# the cost we minimize during training is the NLL of the model
prova=theano.function([x,y],
y,
#updates=updates,
allow_input_downcast=True,on_unused_input='ignore'
)
validate_model=theano.function([x,y],
[full_5_softmax.errors(y),full_5_softmax.y_pred],
allow_input_downcast=True,on_unused_input='ignore'
)
# CARICAMENTO JSON CONTENENTI LE POSELET POSITIVE/NEGATIVE
