def __init__(self):
super(GoogLeNet, self).__init__(
conv1=L.Convolution2D(3, 64, 7, stride=2, pad=3, nobias=True),
norm1=L.BatchNormalization(64),
conv2=L.Convolution2D(64, 192, 3, pad=1, nobias=True),
norm2=L.BatchNormalization(192),
inc3a=L.InceptionBN(192, 64, 64, 64, 64, 96, 'avg', 32),
inc3b=L.InceptionBN(256, 64, 64, 96, 64, 96, 'avg', 64),
inc3c=L.InceptionBN(320, 0, 128, 160, 64, 96, 'max', stride=2),
inc4a=L.InceptionBN(576, 224, 64, 96, 96, 128, 'avg', 128),
inc4b=L.InceptionBN(576, 192, 96, 128, 96, 128, 'avg', 128),
inc4c=L.InceptionBN(576, 128, 128, 160, 128, 160, 'avg', 128),
inc4d=L.InceptionBN(576, 64, 128, 192, 160, 192, 'avg', 128),
inc4e=L.InceptionBN(576, 0, 128, 192, 192, 256, 'max', stride=2),
inc5a=L.InceptionBN(1024, 352, 192, 320, 160, 224, 'avg', 128),
inc5b=L.InceptionBN(1024, 352, 192, 320, 192, 224, 'max', 128),
out_tag=L.Linear(1024 + 8, 3000),
conva=L.Convolution2D(576, 128, 1, nobias=True),
norma=L.BatchNormalization(128),
lina=L.Linear(2048, 1024, nobias=True),
norma2=L.BatchNormalization(1024),
out_a_tag=L.Linear(1024 + 8, 3000),
convb=L.Convolution2D(576, 128, 1, nobias=True),
normb=L.BatchNormalization(128),
linb=L.Linear(2048, 1024, nobias=True),
normb2=L.BatchNormalization(1024),
out_b_tag=L.Linear(1024 + 8, 3000),
)
def __init__(self):
super(GoogLeNetBN, self).__init__(
conv1=L.Convolution2D(None, 64, 7, stride=2, pad=3, nobias=True),
norm1=L.BatchNormalization(64),
conv2=L.Convolution2D(None, 192, 3, pad=1, nobias=True),
norm2=L.BatchNormalization(192),
inc3a=L.InceptionBN(None, 64, 64, 64, 64, 96, 'avg', 32),
inc3b=L.InceptionBN(None, 64, 64, 96, 64, 96, 'avg', 64),
inc3c=L.InceptionBN(None, 0, 128, 160, 64, 96, 'max', stride=2),
inc4a=L.InceptionBN(None, 224, 64, 96, 96, 128, 'avg', 128),
inc4b=L.InceptionBN(None, 192, 96, 128, 96, 128, 'avg', 128),
inc4c=L.InceptionBN(None, 128, 128, 160, 128, 160, 'avg', 128),
inc4d=L.InceptionBN(None, 64, 128, 192, 160, 192, 'avg', 128),
inc4e=L.InceptionBN(None, 0, 128, 192, 192, 256, 'max', stride=2),
inc5a=L.InceptionBN(None, 352, 192, 320, 160, 224, 'avg', 128),
inc5b=L.InceptionBN(None, 352, 192, 320, 192, 224, 'max', 128),
out=L.Linear(None, 1000),
conva=L.Convolution2D(None, 128, 1, nobias=True),
norma=L.BatchNormalization(128),
lina=L.Linear(None, 1024, nobias=True),
norma2=L.BatchNormalization(1024),
outa=L.Linear(None, 1000),
convb=L.Convolution2D(None, 128, 1, nobias=True),
normb=L.BatchNormalization(128),
linb=L.Linear(None, 1024, nobias=True),
normb2=L.BatchNormalization(1024),
outb=L.Linear(None, 1000),
)
self._train = True
def __init__(self, n_classes, n_base_units=8):
n_mid_units = n_base_units * 2
super().__init__(
# conv1: 100 -> (100 - 4) / 3 + 1 = 33
conv1=L.Convolution2D(3, n_base_units, 4, stride=3),
# pool1: 33 -> (33 + 1*2 - 3) / 2 + 1 = 17
# inc1: 17 -> (17 + 1*2 - 3) / 2 + 1 = 9
inc1=L.InceptionBN(in_channels=n_base_units,
out1=0,
proj3=n_mid_units,
out3=n_mid_units,
proj33=n_mid_units,
out33=n_mid_units,
pooltype='max',
stride=2),
# inc2: 9 -> (9 + 1*2 - 3) / 2 + 1 = 5
inc2=L.InceptionBN(in_channels=n_mid_units * 2 + n_base_units,
out1=0,
proj3=n_mid_units,
out3=n_mid_units,
proj33=n_mid_units,
out33=n_mid_units,
pooltype='max',
stride=2),
fc1=L.Linear(
5 * 5 * (n_mid_units * 2 + n_mid_units * 2 + n_base_units),
128),
fc2=L.Linear(128, n_classes))
self.n_units = 5 * 5 * (n_mid_units * 2 + n_mid_units * 2 +
n_base_units)
self.train = True
def __init__(self):
self.dtype = dtype = np.float16
W = initializers.HeNormal(1 / np.sqrt(2), self.dtype)
bias = initializers.Zero(self.dtype)
chainer.Chain.__init__(self)
with self.init_scope():
self.conv1 = L.Convolution2D(
None, 64, 7, stride=2, pad=3, initialW=W, nobias=True)
self.norm1 = L.BatchNormalization(64, dtype=dtype)
self.conv2 = L.Convolution2D(
None, 192, 3, pad=1, initialW=W, nobias=True)
self.norm2 = L.BatchNormalization(192, dtype=dtype)
self.inc3a = L.InceptionBN(
None, 64, 64, 64, 64, 96, 'avg', 32, conv_init=W, dtype=dtype)
self.inc3b = L.InceptionBN(
None, 64, 64, 96, 64, 96, 'avg', 64, conv_init=W, dtype=dtype)
self.inc3c = L.InceptionBN(
None, 0, 128, 160, 64, 96, 'max', stride=2,
conv_init=W, dtype=dtype)
self.inc4a = L.InceptionBN(
None, 224, 64, 96, 96, 128, 'avg', 128,
conv_init=W, dtype=dtype)
self.inc4b = L.InceptionBN(
None, 192, 96, 128, 96, 128, 'avg', 128,
conv_init=W, dtype=dtype)
self.inc4c = L.InceptionBN(
None, 128, 128, 160, 128, 160, 'avg', 128,
conv_init=W, dtype=dtype)
self.inc4d = L.InceptionBN(
None, 64, 128, 192, 160, 192, 'avg', 128,
conv_init=W, dtype=dtype)
self.inc4e = L.InceptionBN(
None, 0, 128, 192, 192, 256, 'max',
stride=2, conv_init=W, dtype=dtype)
self.inc5a = L.InceptionBN(
None, 352, 192, 320, 160, 224, 'avg', 128,
conv_init=W, dtype=dtype)
self.inc5b = L.InceptionBN(
None, 352, 192, 320, 192, 224, 'max', 128,
conv_init=W, dtype=dtype)
self.out = L.Linear(None, 1000, initialW=W, bias=bias)
self.conva = L.Convolution2D(None, 128, 1, initialW=W, nobias=True)
self.norma = L.BatchNormalization(128, dtype=dtype)
self.lina = L.Linear(None, 1024, initialW=W, nobias=True)
self.norma2 = L.BatchNormalization(1024, dtype=dtype)
self.outa = L.Linear(None, 1000, initialW=W, bias=bias)
self.convb = L.Convolution2D(None, 128, 1, initialW=W, nobias=True)
self.normb = L.BatchNormalization(128, dtype=dtype)
self.linb = L.Linear(None, 1024, initialW=W, nobias=True)
self.normb2 = L.BatchNormalization(1024, dtype=dtype)
self.outb = L.Linear(None, 1000, initialW=W, bias=bias)
def __init__(self, n_class):
super(GoogLeNetBN, self).__init__()
with self.init_scope():
self.conv1 = L.Convolution2D(None,
64,
7,
stride=2,
pad=3,
nobias=True)
self.norm1 = L.BatchNormalization(64)
self.conv2 = L.Convolution2D(None, 192, 3, pad=1, nobias=True)
self.norm2 = L.BatchNormalization(192)
self.inc3a = L.InceptionBN(None, 64, 64, 64, 64, 96, 'avg', 32)
self.inc3b = L.InceptionBN(None, 64, 64, 96, 64, 96, 'avg', 64)
self.inc3c = L.InceptionBN(None,
0,
128,
160,
64,
96,
'max',
stride=2)
self.inc4a = L.InceptionBN(None, 224, 64, 96, 96, 128, 'avg', 128)
self.inc4b = L.InceptionBN(None, 192, 96, 128, 96, 128, 'avg', 128)
self.inc4c = L.InceptionBN(None, 160, 128, 160, 128, 160, 'avg',
128)
self.inc4d = L.InceptionBN(None, 96, 128, 192, 160, 192, 'avg',
128)
self.inc4e = L.InceptionBN(None,
0,
128,
192,
192,
256,
'max',
stride=2)
self.inc5a = L.InceptionBN(None, 352, 192, 320, 160, 224, 'avg',
128)
self.inc5b = L.InceptionBN(None, 352, 192, 320, 192, 224, 'max',
128)
self.out = L.Linear(None, n_class)
self.conva = L.Convolution2D(None, 128, 1, nobias=True)
self.norma = L.BatchNormalization(128)
self.lina = L.Linear(None, 1024, nobias=True)
self.norma2 = L.BatchNormalization(1024)
self.outa = L.Linear(None, n_class)
self.convb = L.Convolution2D(None, 128, 1, nobias=True)
self.normb = L.BatchNormalization(128)
self.linb = L.Linear(None, 1024, nobias=True)
self.normb2 = L.BatchNormalization(1024)
self.outb = L.Linear(None, n_class)
def setup_data(self):
dtype = chainer.get_dtype()
self.x = numpy.random.uniform(
-1, 1, (10, self.in_channels, 5, 5)).astype(dtype)
self.l = links.InceptionBN(self.in_channels, self.out1, self.proj3,
self.out3, self.proj33, self.out33,
self.pooltype, self.proj_pool, self.stride)
def setup_data(self):
self.x = numpy.random.uniform(
-1, 1, (10, self.in_channels, 5, 5)
).astype(numpy.float32)
self.l = links.InceptionBN(
self.in_channels, self.out1, self.proj3, self.out3,
self.proj33, self.out33, self.pooltype, self.proj_pool,
self.stride)
def __init__(self):
super(CNN_02, self).__init__(
conv1=L.Convolution2D(1, 64, 3, stride=1, nobias=True),
# MaxPool(3x3, 2)
bn1=L.BatchNormalization(64),
conv2a=L.Convolution2D(64, 64, 1, stride=1),
conv2b=L.Convolution2D(64, 128, 3, stride=1),
bn2=L.BatchNormalization(128),
# MaxPool(3x3, 2)
inc3a=L.InceptionBN(128, 64, 48, 64, 64, 96, 'avg', 32),
inc3b=L.InceptionBN(256, 64, 64, 96, 64, 96, 'max', 64),
# MaxPool(3x3, 2)
inc4a=L.InceptionBN(320, 224, 64, 96, 96, 128, 'avg', 128),
inc4b=L.InceptionBN(576, 192, 128, 192, 192, 256, 'max', 128),
inc5a=L.InceptionBN(768, 320, 192, 288, 192, 288, 'max', 128),
# AveragePool(7x7, 1)
# Dropout(40%)
preout=L.Linear(1024, 64),
# ReLu
out=L.Linear(64, 2)
# SoftMax
)
def __init__(self, n_outputs):
super(GoogLeNetBN, self).__init__(
conv1=L.Convolution2D(3, 64, 7, stride=2, pad=3, nobias=True),
norm1=L.BatchNormalization(64),
conv2=L.Convolution2D(64, 192, 3, pad=1, nobias=True),
norm2=L.BatchNormalization(192),
inc3a=L.InceptionBN(192, 64, 64, 64, 64, 96, 'avg', 32),
inc3b=L.InceptionBN(256, 64, 64, 96, 64, 96, 'avg', 64),
inc3c=L.InceptionBN(320, 0, 128, 160, 64, 96, 'max', stride=2),
inc4a=L.InceptionBN(576, 224, 64, 96, 96, 128, 'avg', 128),
inc4b=L.InceptionBN(576, 192, 96, 128, 96, 128, 'avg', 128),
inc4c=L.InceptionBN(576, 128, 128, 160, 128, 160, 'avg', 128),
inc4d=L.InceptionBN(576, 64, 128, 192, 160, 192, 'avg', 128),
inc4e=L.InceptionBN(576, 0, 128, 192, 192, 256, 'max', stride=2),
inc5a=L.InceptionBN(1024, 352, 192, 320, 160, 224, 'avg', 128),
inc5b=L.InceptionBN(1024, 352, 192, 320, 192, 224, 'max', 128),
linz=L.Linear(1024, 300),
out=L.Linear(300, n_outputs),
outimg=L.Linear(1024, n_outputs),
outdoc=L.Linear(1000, n_outputs),
doc_fc1=L.Linear(1000, 600),
doc_fc2=L.Linear(600, 300),
conva=L.Convolution2D(576, 128, 1, nobias=True),
norma=L.BatchNormalization(128),
lina=L.Linear(3200, 1024, nobias=True),
norma2=L.BatchNormalization(1024),
outa=L.Linear(1024, n_outputs),
convb=L.Convolution2D(576, 128, 1, nobias=True),
normb=L.BatchNormalization(128),
linb=L.Linear(3200, 1024, nobias=True),
normb2=L.BatchNormalization(1024),
outb=L.Linear(1024, n_outputs),
bi1=L.Bilinear(300, 300, 300)
)
def __init__(self,
n_class=None,
pretrained_model=None,
mean=None,
initialW=None,
initialBias=None):
self.n_class = n_class
self.mean = mean
self.initialbias = initialBias
self.insize = 224
if n_class is None:
self.n_class = 100
if mean is None:
# imagenet means
self.mean = np.array([123.68, 116.779, 103.939],
dtype=np.float32)[:, np.newaxis, np.newaxis]
if initialW is None:
# employ default initializers used in BVLC. For more detail, see
self.initialW = uniform.LeCunUniform(scale=1.0)
if pretrained_model is None:
# As a sampling process is time-consuming
# we employ a zero initializer for faster computation
self.initialW = constant.Zero()
super(GoogleNetBN, self).__init__()
with self.init_scope():
# Deep layers: GoogleNet of BatchNormalization version
self.conv1 = L.Convolution2D(None,
64,
7,
stride=2,
pad=3,
nobias=True)
self.norm1 = L.BatchNormalization(64)
self.conv2 = L.Convolution2D(None,
192,
3,
stride=1,
pad=1,
nobias=True)
self.norm2 = L.BatchNormalization(192)
self.inc3a = L.InceptionBN(None, 64, 64, 64, 64, 96, "avg", 32)
self.inc3b = L.InceptionBN(None, 64, 64, 96, 64, 96, "avg", 64)
self.inc3c = L.InceptionBN(None,
0,
128,
160,
64,
96,
"max",
stride=2)
self.inc4a = L.InceptionBN(None, 224, 64, 96, 96, 128, "avg", 128)
self.inc4b = L.InceptionBN(None, 192, 96, 128, 96, 128, "avg", 128)
self.inc4c = L.InceptionBN(None, 128, 128, 160, 128, 160, "avg",
128)
self.inc4d = L.InceptionBN(None, 64, 128, 192, 160, 192, "avg",
128)
self.inc4e = L.InceptionBN(None,
0,
128,
192,
192,
256,
"max",
stride=2)
self.inc5a = L.InceptionBN(None, 352, 192, 320, 160, 224, "avg",
128)
self.inc5b = L.InceptionBN(None, 352, 192, 320, 192, 224, "max",
128)
self.loss3_fc = L.Linear(None,
self.n_class,
initialW=self.initialW)
self.loss1_conv = L.Convolution2D(None,
128,
1,
initialW=self.initialW,
nobias=True)
self.norma = L.BatchNormalization(128)
self.loss1_fc1 = L.Linear(None,
1024,
initialW=self.initialW,
nobias=True)
self.norma2 = L.BatchNormalization(1024)
self.loss1_fc2 = L.Linear(None,
self.n_class,
initialW=self.initialW)
self.loss2_conv = L.Convolution2D(None,
128,
1,
initialW=self.initialW,
nobias=True)
self.normb = L.BatchNormalization(128)
self.loss2_fc1 = L.Linear(None,
1024,
initialW=self.initialW,
nobias=True)
self.normb2 = L.BatchNormalization(1024)
self.loss2_fc2 = L.Linear(None,
self.n_class,
initialW=self.initialW)
