def __init__(self, n_out):
super(GoogLeNet, self).__init__(
conv1=L.Convolution2D(None, 64, 7, stride=1, pad=3),
conv2_reduce=L.Convolution2D(None, 64, 1),
conv2=L.Convolution2D(None, 192, 3, stride=1, pad=1),
inc3a=L.Inception(None, 64, 96, 128, 16, 32, 32),
inc3b=L.Inception(None, 128, 128, 192, 32, 96, 64),
inc4a=L.Inception(None, 192, 96, 208, 16, 48, 64),
inc4b=L.Inception(None, 160, 112, 224, 24, 64, 64),
inc4c=L.Inception(None, 128, 128, 256, 24, 64, 64),
inc4d=L.Inception(None, 112, 144, 288, 32, 64, 64),
inc4e=L.Inception(None, 256, 160, 320, 32, 128, 128),
inc5a=L.Inception(None, 256, 160, 320, 32, 128, 128),
inc5b=L.Inception(None, 384, 192, 384, 48, 128, 128),
loss3_fc=L.Linear(None, n_out),
loss1_conv=L.Convolution2D(None, 128, 1),
loss1_fc1=L.Linear(None, 1024),
loss1_fc2=L.Linear(None, n_out),
loss2_conv=L.Convolution2D(None, 128, 1),
loss2_fc1=L.Linear(None, 1024),
loss2_fc2=L.Linear(None, n_out)
)
self.train = True
def __init__(self, num_class=1000, texture=False, cbp=False, normalize=True, mil=False):
super(GoogLeNet, self).__init__()
with self.init_scope():
self.conv1 = L.Convolution2D(None, 64, 7, stride=2, pad=3)
self.conv2_reduce = L.Convolution2D(None, 64, 1)
self.conv2 = L.Convolution2D(None, 192, 3, stride=1, pad=1)
self.inc3a = L.Inception(None, 64, 96, 128, 16, 32, 32)
self.inc3b = L.Inception(None, 128, 128, 192, 32, 96, 64)
self.inc4a = L.Inception(None, 192, 96, 208, 16, 48, 64)
self.inc4b = L.Inception(None, 160, 112, 224, 24, 64, 64)
self.inc4c = L.Inception(None, 128, 128, 256, 24, 64, 64)
self.inc4d = L.Inception(None, 112, 144, 288, 32, 64, 64)
if not texture:
self.inc4e = L.Inception(None, 256, 160, 320, 32, 128, 128)
self.inc5a = L.Inception(None, 256, 160, 320, 32, 128, 128)
self.inc5b = L.Inception(None, 384, 192, 384, 48, 128, 128)
self.loss1_conv = L.Convolution2D(None, 128, 1)
self.loss1_fc1 = L.Linear(None, 1024)
self.loss1_fc2 = L.Linear(None, num_class)
self.loss2_conv = L.Convolution2D(None, 128, 1)
self.loss2_fc1 = L.Linear(None, 1024)
self.loss2_fc2 = L.Linear(None, num_class)
self.loss3_fc = L.Linear(None, num_class)
self.texture = texture
self.cbp = cbp
self.normalize = normalize
self.mil = mil
def __init__(self):
super(Network,
self).__init__(conv1=L.Convolution2D(3, 64, 7, stride=2, pad=3),
conv2_reduce=L.Convolution2D(64, 64, 1),
conv2=L.Convolution2D(64, 192, 3, stride=1,
pad=1),
inc3a=L.Inception(192, 64, 96, 128, 16, 32, 32),
inc3b=L.Inception(256, 128, 128, 192, 32, 96, 64),
inc4a=L.Inception(480, 192, 96, 208, 16, 48, 64),
inc4b=L.Inception(512, 160, 112, 224, 24, 64, 64),
inc4c=L.Inception(512, 128, 128, 256, 24, 64, 64),
inc4d=L.Inception(512, 112, 144, 288, 32, 64, 64),
inc4e=L.Inception(528, 256, 160, 320, 32, 128,
128),
inc5a=L.Inception(832, 256, 160, 320, 32, 128,
128),
inc5b=L.Inception(832, 384, 192, 384, 48, 128,
128),
loss3_fc=L.Linear(1024, 1000),
loss1_conv=L.Convolution2D(512, 128, 1),
loss1_fc1=L.Linear(4 * 4 * 128, 1024),
loss1_fc2=L.Linear(1024, 1000),
loss2_conv=L.Convolution2D(528, 128, 1),
loss2_fc1=L.Linear(4 * 4 * 128, 1024),
loss2_fc2=L.Linear(1024, 1000))
self.train = True
def __init__(self):
super(GoogLeNet, self).__init__(
conv1=L.Convolution2D(3, 64, 7, stride=2, pad=3),
conv2_reduce=L.Convolution2D(64, 64, 1),
conv2=L.Convolution2D(64, 192, 3, stride=1, pad=1),
#inception consists of 1*1,3*3,5*5convolution and maxpooling(convolution)
#http://docs.chainer.org/en/stable/_modules/chainer/links/connection/inception.html?highlight=Inception
#inchannel, out1,proj3,out3,proj5,out5,proj_pool,conv_init=none, biase_init=none
inc3a=L.Inception(192, 64, 96, 128, 16, 32, 32),
inc3b=L.Inception(256, 128, 128, 192, 32, 96, 64),
inc4a=L.Inception(480, 192, 96, 208, 16, 48, 64),
inc4b=L.Inception(512, 160, 112, 224, 24, 64, 64),
inc4c=L.Inception(512, 128, 128, 256, 24, 64, 64),
inc4d=L.Inception(512, 112, 144, 288, 32, 64, 64),
inc4e=L.Inception(528, 256, 160, 320, 32, 128, 128),
inc5a=L.Inception(832, 256, 160, 320, 32, 128, 128),
inc5b=L.Inception(832, 384, 192, 384, 48, 128, 128),
loss3_fc=L.Linear(1024, 1000),
loss1_conv=L.Convolution2D(512, 128, 1),
loss1_fc1=L.Linear(4 * 4 * 128, 1024),
loss1_fc2=L.Linear(1024, 1000),
loss2_conv=L.Convolution2D(528, 128, 1),
loss2_fc1=L.Linear(4 * 4 * 128, 1024),
loss2_fc2=L.Linear(1024, 1000))
self.train = True
def __init__(self, output_size=1000):
super(GoogLeNet, self).__init__()
with self.init_scope():
self.conv1 = L.Convolution2D(3, 64, 7, stride=2, pad=3)
self.conv2_reduce = L.Convolution2D(64, 64, 1)
self.conv2 = L.Convolution2D(64, 192, 3, stride=1, pad=1)
self.inc3a = L.Inception(192, 64, 96, 128, 16, 32, 32)
self.inc3b = L.Inception(256, 128, 128, 192, 32, 96, 64)
self.inc4a = L.Inception(480, 192, 96, 208, 16, 48, 64)
self.inc4b = L.Inception(512, 160, 112, 224, 24, 64, 64)
self.inc4c = L.Inception(512, 128, 128, 256, 24, 64, 64)
self.inc4d = L.Inception(512, 112, 144, 288, 32, 64, 64)
self.inc4e = L.Inception(528, 256, 160, 320, 32, 128, 128)
self.inc5a = L.Inception(832, 256, 160, 320, 32, 128, 128)
self.inc5b = L.Inception(832, 384, 192, 384, 48, 128, 128)
self.loss3_fc = L.Linear(1024, 256)
self.append_loss3_fc = L.Linear(256, output_size)
self.loss1_conv = L.Convolution2D(512, 128, 1)
self.loss1_fc1 = L.Linear(2048, 1024)
self.loss1_fc2 = L.Linear(1024, output_size)
self.loss2_conv = L.Convolution2D(528, 128, 1)
self.loss2_fc1 = L.Linear(2048, 1024)
self.loss2_fc2 = L.Linear(1024, output_size)
def __init__(self, n_outputs):
super(GoogLeNet, self).__init__(
conv1=L.Convolution2D(3, 64, 7, stride=2, pad=3),
conv2_reduce=L.Convolution2D(64, 64, 1),
conv2=L.Convolution2D(64, 192, 3, stride=1, pad=1),
inc3a=L.Inception(192, 64, 96, 128, 16, 32, 32),
inc3b=L.Inception(256, 128, 128, 192, 32, 96, 64),
inc4a=L.Inception(480, 192, 96, 208, 16, 48, 64),
inc4b=L.Inception(512, 160, 112, 224, 24, 64, 64),
inc4c=L.Inception(512, 128, 128, 256, 24, 64, 64),
inc4d=L.Inception(512, 112, 144, 288, 32, 64, 64),
inc4e=L.Inception(528, 256, 160, 320, 32, 128, 128),
inc5a=L.Inception(832, 256, 160, 320, 32, 128, 128),
inc5b=L.Inception(832, 384, 192, 384, 48, 128, 128),
loss3_fc1=L.Linear(4096, 300),
loss3_fc2=L.Linear(300, n_outputs),
doc_fc1=L.Linear(1000, 600),
doc_fc2=L.Linear(600, 300),
bi1=L.Bilinear(300, 300, 300),
loss1_conv=L.Convolution2D(512, 128, 1),
loss1_fc1=L.Linear(4 * 4 * 128, 1024),
loss1_fc2=L.Linear(1024, n_outputs),
loss2_conv=L.Convolution2D(528, 128, 1),
loss2_fc1=L.Linear(4 * 4 * 128, 1024),
loss2_fc2=L.Linear(1024, n_outputs),
)
def setUp(self):
self.x = numpy.random.uniform(
-1, 1, (10, self.in_channels, 5, 5)).astype(numpy.float32)
out = self.out1 + self.out3 + self.out5 + self.proj_pool
self.gy = numpy.random.uniform(-1, 1,
(10, out, 5, 5)).astype(numpy.float32)
self.l = links.Inception(self.in_channels, self.out1, self.proj3,
self.out3, self.proj5, self.out5,
self.proj_pool)
def __init__(self):
super(GoogLeNet, self).__init__()
with self.init_scope():
self.conv1 = L.Convolution2D(None, 64, 7, stride=2, pad=3)
self.conv2_reduce = L.Convolution2D(None, 64, 1)
self.conv2 = L.Convolution2D(None, 192, 3, stride=1, pad=1)
self.inc3a = L.Inception(None, 64, 96, 128, 16, 32, 32)
self.inc3b = L.Inception(None, 128, 128, 192, 32, 96, 64)
self.inc4a = L.Inception(None, 192, 96, 208, 16, 48, 64)
self.inc4b = L.Inception(None, 160, 112, 224, 24, 64, 64)
self.inc4c = L.Inception(None, 128, 128, 256, 24, 64, 64)
self.inc4d = L.Inception(None, 112, 144, 288, 32, 64, 64)
self.inc4e = L.Inception(None, 256, 160, 320, 32, 128, 128)
self.inc5a = L.Inception(None, 256, 160, 320, 32, 128, 128)
self.inc5b = L.Inception(None, 384, 192, 384, 48, 128, 128)
self.loss3_fc = L.Linear(None, 1000)
self.loss1_conv = L.Convolution2D(None, 128, 1)
self.loss1_fc1 = L.Linear(None, 1024)
self.loss1_fc2 = L.Linear(None, 1000)
self.loss2_conv = L.Convolution2D(None, 128, 1)
self.loss2_fc1 = L.Linear(None, 1024)
self.loss2_fc2 = L.Linear(None, 1000)
batch = 1
total_backward = 0
total_forward = 0
count = 0
niter = 5
n_dry = 3
# data = np.ndarray((batch, 3, 224, 224), dtype=np.float32)
# data.fill(333.33)
data = np.random.rand(batch, 8, 3, 3).astype(np.float32)
y_grad = np.ones((batch, 32, 3, 3), dtype=np.float32)
# incept = L.Inception(3, 64, 96, 128, 16, 32, 32)
incept = L.Inception(8, 8, 96, 8, 16, 8, 8)
x = Variable(data)
for i in range(niter):
print("iter:", i)
start = time.time()
y = incept(x)
end = time.time()
if i > n_dry - 1:
count += 1
total_forward += (end - start) * 1000
y.grad = y_grad
start = time.time()
y.backward()
end = time.time()
