def forwardSize(self, inputsize):
# [size1, size2, size3], size: (32,1,28,28)
# print("conv2d.size: {}, {}, {}".format(inputsize,self.mapMulti, self.inputFeature))
isize = inputsize[0]
if len(isize) != 4:
raise IndexError
if self.mapMulti is None and isize[1] != self.inputFeature:
raise IndexError
if self.mapMulti is not None:
self.inputFeature = isize[1]
self.outputFeature = int(self.inputFeature * self.mapMulti)
weightIniter = winit.XavierInit()
initweight = weightIniter.initialize(
(self.outputFeature, self.inputFeature, *self.filterSize))
initbias = np.zeros((self.outputFeature, ))
self.w = theano.shared(initweight, borrow=True)
self.b = theano.shared(initbias, borrow=True)
retSize = None
if self.border == 'half':
retSize = [(isize[0], self.outputFeature,
int(isize[2] / self.subsample[0]),
int(isize[3] / self.subsample[1]))]
else:
raise NotImplementedError
return retSize
def forwardSize(self, inputsize):
isize = inputsize[0]
if len(isize) != 4:
raise AssertionError('Conv input size should has a lenght of 4')
if self.inputFeature is not None and isize[1] != self.inputFeature:
raise AssertionError('Conv put feature map size does not match')
if self.outputFeature is not None:
self.inputFeature = isize[1]
self.mapMulti = self.outputFeature / self.inputFeature
elif self.mapMulti is not None:
self.inputFeature = isize[1]
self.outputFeature = int(self.inputFeature * self.mapMulti)
weightIniter = winit.XavierInit()
initweight = weightIniter.initialize(
(self.outputFeature, self.inputFeature, *self.filterSize))
initbias = floatX(np.zeros((self.outputFeature, )))
self.w = theano.shared(initweight, borrow=True)
self.b = theano.shared(initbias, borrow=True)
retSize = None
if self.border == 'half':
retSize = [(isize[0], self.outputFeature,
int(isize[2] / self.subsample[0]),
int(isize[3] / self.subsample[1]))]
else:
raise NotImplementedError
return retSize
def forwardSize(self, inputsize):
#print(inputsize)
#print(self.inputFeature)
isize = inputsize[0]
if len(isize) != 2:
raise IndexError('Expect input dimension 2, get ' +
str(len(isize)))
if self.inputFeature is not None and isize[1] != self.inputFeature:
raise IndexError('Input size: ' + str(isize[1]) +
' is not equal to given input feature dim: ' +
str(self.inputFeature))
if self.inputFeature is None:
self.inputFeature = isize[1]
if self.outputFeature is None:
self.outputFeature = self.inputFeature * self.mapMulti
weightIniter = winit.XavierInit()
initweight = weightIniter.initialize(
(self.inputFeature, self.outputFeature))
self.w = theano.shared(initweight, borrow=True)
initbias = np.zeros((self.outputFeature, ))
self.b = theano.shared(initbias, borrow=True)
return [(
isize[0],
self.outputFeature,
)]
def __init__(self, times=None, output=None, input_feature=None, output_feature=None):
super(FullConn, self).__init__()
if times is not None:
self.times = times
if output is not None:
self.output = output
weightIniter = winit.XavierInit()
initweight = weightIniter.initialize((input_feature, output_feature))
self.w = theano.shared(initweight, borrow=True)
self.inputFeature = input_feature
self.outputFeature = output_feature
self.times = -1
self.output = -1