def __init__(self, n_classes=6):
super(SteerCNN, self).__init__()
# the model is equivariant under rotations by 45 degrees, modelled by C8
self.r2_act = gspaces.Rot2dOnR2(N=4)
# the input image is a scalar field, corresponding to the trivial representation
input_type = nn_e2.FieldType(self.r2_act,
3 * [self.r2_act.trivial_repr])
# we store the input type for wrapping the images into a geometric tensor during the forward pass
self.input_type = input_type
# convolution 1
# first specify the output type of the convolutional layer
# we choose 24 feature fields, each transforming under the regular representation of C8
out_type = nn_e2.FieldType(self.r2_act,
24 * [self.r2_act.regular_repr])
self.block1 = nn_e2.SequentialModule(
nn_e2.R2Conv(input_type,
out_type,
kernel_size=7,
padding=3,
bias=False), nn_e2.InnerBatchNorm(out_type),
nn_e2.ReLU(out_type, inplace=True))
self.pool1 = nn_e2.PointwiseAvgPool(out_type, 4)
# convolution 2
# the old output type is the input type to the next layer
in_type = self.block1.out_type
# the output type of the second convolution layer are 48 regular feature fields of C8
#out_type = nn_e2.FieldType(self.r2_act, 48 * [self.r2_act.regular_repr])
self.block2 = nn_e2.SequentialModule(
nn_e2.R2Conv(in_type,
out_type,
kernel_size=7,
padding=3,
bias=False), nn_e2.InnerBatchNorm(out_type),
nn_e2.ReLU(out_type, inplace=True))
self.pool2 = nn_e2.SequentialModule(
nn_e2.PointwiseAvgPoolAntialiased(out_type,
sigma=0.66,
stride=1,
padding=0),
nn_e2.PointwiseAvgPool(out_type, 4), nn_e2.GroupPooling(out_type))
# PointwiseAvgPoolAntialiased(out_type, sigma=0.66, stride=7)
# number of output channels
c = 24 * 13 * 13 #self.gpool.out_type.size
# Fully Connected
self.fully_net = torch.nn.Sequential(
torch.nn.Linear(c, 64),
torch.nn.BatchNorm1d(64),
torch.nn.ELU(inplace=True),
torch.nn.Linear(64, n_classes),
)
def __init__(self, channel_in=3, n_classes=4, rot_n=4):
super(SmallE2, self).__init__()
r2_act = gspaces.Rot2dOnR2(N=rot_n)
self.feat_type_in = nn.FieldType(r2_act,
channel_in * [r2_act.trivial_repr])
feat_type_hid = nn.FieldType(r2_act, 8 * [r2_act.regular_repr])
feat_type_out = nn.FieldType(r2_act, 2 * [r2_act.regular_repr])
self.bn = nn.InnerBatchNorm(feat_type_hid)
self.relu = nn.ReLU(feat_type_hid)
self.convin = nn.R2Conv(self.feat_type_in,
feat_type_hid,
kernel_size=3)
self.convhid = nn.R2Conv(feat_type_hid, feat_type_hid, kernel_size=3)
self.convout = nn.R2Conv(feat_type_hid, feat_type_out, kernel_size=3)
self.avgpool = nn.PointwiseAvgPool(feat_type_out, 3)
self.invariant_map = nn.GroupPooling(feat_type_out)
c = self.invariant_map.out_type.size
self.lin_in = torch.nn.Linear(c, 64)
self.elu = torch.nn.ELU()
self.lin_out = torch.nn.Linear(64, n_classes)
def __init__(self):
super(DenseFeatureExtractionModuleE2Inv, self).__init__()
filters = np.array([32,32, 64,64, 128,128,128, 256,256,256, 512,512,512], dtype=np.int32)*2
# number of rotations to consider for rotation invariance
N = 8
self.gspace = gspaces.Rot2dOnR2(N)
self.input_type = enn.FieldType(self.gspace, [self.gspace.trivial_repr] * 3)
ip_op_types = [
self.input_type,
]
self.num_channels = 64
for filter_ in filters[:10]:
ip_op_types.append(FIELD_TYPE['regular'](self.gspace, filter_, fixparams=False))
self.model = enn.SequentialModule(*[
conv3x3(ip_op_types[0], ip_op_types[1]),
enn.ReLU(ip_op_types[1], inplace=True),
conv3x3(ip_op_types[1], ip_op_types[2]),
enn.ReLU(ip_op_types[2], inplace=True),
enn.PointwiseMaxPool(ip_op_types[2], 2),
conv3x3(ip_op_types[2], ip_op_types[3]),
enn.ReLU(ip_op_types[3], inplace=True),
conv3x3(ip_op_types[3], ip_op_types[4]),
enn.ReLU(ip_op_types[4], inplace=True),
enn.PointwiseMaxPool(ip_op_types[4], 2),
conv3x3(ip_op_types[4], ip_op_types[5]),
enn.ReLU(ip_op_types[5], inplace=True),
conv3x3(ip_op_types[5], ip_op_types[6]),
enn.ReLU(ip_op_types[6], inplace=True),
conv3x3(ip_op_types[6], ip_op_types[7]),
enn.ReLU(ip_op_types[7], inplace=True),
enn.PointwiseAvgPool(ip_op_types[7], kernel_size=2, stride=1),
conv5x5(ip_op_types[7], ip_op_types[8]),
enn.ReLU(ip_op_types[8], inplace=True),
conv5x5(ip_op_types[8], ip_op_types[9]),
enn.ReLU(ip_op_types[9], inplace=True),
conv5x5(ip_op_types[9], ip_op_types[10]),
enn.ReLU(ip_op_types[10], inplace=True),
# enn.PointwiseMaxPool(ip_op_types[7], 2),
# conv3x3(ip_op_types[7], ip_op_types[8]),
# enn.ReLU(ip_op_types[8], inplace=True),
# conv3x3(ip_op_types[8], ip_op_types[9]),
# enn.ReLU(ip_op_types[9], inplace=True),
# conv3x3(ip_op_types[9], ip_op_types[10]),
# enn.ReLU(ip_op_types[10], inplace=True),
enn.GroupPooling(ip_op_types[10])
])
def __init__(self, in_type, out_type, gspace):
super(TransitionBlock, self).__init__()
self.gspace = gspace
self.in_type = FIELD_TYPE["regular"](self.gspace, in_type, fixparams=False)
self.out_type = FIELD_TYPE["regular"](self.gspace, out_type, fixparams=False)
self.bn1 = enn.InnerBatchNorm(self.in_type)
self.relu1 = enn.ReLU(self.in_type,inplace=True)
self.conv1 = conv1x1(self.in_type,self.out_type)
self.avgpool = enn.PointwiseAvgPool(self.out_type, kernel_size=2)
def ennAvgPool(inplanes,
kernel_size=1,
stride=None,
padding=0,
ceil_mode=False):
in_type = FIELD_TYPE['regular'](gspace, inplanes)
return enn.PointwiseAvgPool(in_type,
kernel_size,
stride=stride,
padding=padding,
ceil_mode=ceil_mode)
def __init__(self, n_classes=10):
super(C8SteerableCNN, self).__init__()
# the model is equivariant under rotations by 45 degrees, modelled by C8
self.r2_act = gspaces.Rot2dOnR2(N=8)
# the input image is a scalar field, corresponding to the trivial representation
in_type = nn.FieldType(self.r2_act, [self.r2_act.trivial_repr])
# we store the input type for wrapping the images into a geometric tensor during the forward pass
self.input_type = in_type
# convolution 1
# first specify the output type of the convolutional layer
# we choose 16 feature fields, each transforming under the regular representation of C8
out_type = nn.FieldType(self.r2_act, 24 * [self.r2_act.regular_repr])
self.block1 = nn.SequentialModule(
# nn.MaskModule(in_type, 29, margin=1),
nn.R2Conv(in_type, out_type, kernel_size=7, padding=1, bias=False),
nn.InnerBatchNorm(out_type),
nn.ReLU(out_type, inplace=True))
# convolution 2
# the old output type is the input type to the next layer
in_type = self.block1.out_type
# the output type of the second convolution layer are 32 regular feature fields of C8
out_type = nn.FieldType(self.r2_act, 48 * [self.r2_act.regular_repr])
self.block2 = nn.SequentialModule(
nn.R2Conv(in_type, out_type, kernel_size=5, padding=2, bias=False),
nn.InnerBatchNorm(out_type), nn.ReLU(out_type, inplace=True))
self.pool1 = nn.SequentialModule(
nn.PointwiseAvgPoolAntialiased(out_type, sigma=0.66, stride=2))
# convolution 3
# the old output type is the input type to the next layer
in_type = self.block2.out_type
# the output type of the third convolution layer are 32 regular feature fields of C8
out_type = nn.FieldType(self.r2_act, 48 * [self.r2_act.regular_repr])
self.block3 = nn.SequentialModule(
nn.R2Conv(in_type, out_type, kernel_size=5, padding=2, bias=False),
nn.InnerBatchNorm(out_type), nn.ReLU(out_type, inplace=True))
# convolution 4
# the old output type is the input type to the next layer
in_type = self.block3.out_type
# the output type of the fourth convolution layer are 64 regular feature fields of C8
out_type = nn.FieldType(self.r2_act, 96 * [self.r2_act.regular_repr])
self.block4 = nn.SequentialModule(
nn.R2Conv(in_type, out_type, kernel_size=5, padding=2, bias=False),
nn.InnerBatchNorm(out_type), nn.ReLU(out_type, inplace=True))
self.pool2 = nn.SequentialModule(
nn.PointwiseAvgPoolAntialiased(out_type, sigma=0.66, stride=2))
# convolution 5
# the old output type is the input type to the next layer
in_type = self.block4.out_type
# the output type of the fifth convolution layer are 64 regular feature fields of C8
out_type = nn.FieldType(self.r2_act, 96 * [self.r2_act.regular_repr])
self.block5 = nn.SequentialModule(
nn.R2Conv(in_type, out_type, kernel_size=5, padding=2, bias=False),
nn.InnerBatchNorm(out_type), nn.ReLU(out_type, inplace=True))
# convolution 6
# the old output type is the input type to the next layer
in_type = self.block5.out_type
# the output type of the sixth convolution layer are 64 regular feature fields of C8
out_type = nn.FieldType(self.r2_act, 64 * [self.r2_act.regular_repr])
self.block6 = nn.SequentialModule(
nn.R2Conv(in_type, out_type, kernel_size=5, padding=1, bias=False),
nn.InnerBatchNorm(out_type), nn.ReLU(out_type, inplace=True))
self.pool3 = nn.PointwiseAvgPool(out_type, kernel_size=4)
self.gpool = nn.GroupPooling(out_type)
# number of output channels
c = self.gpool.out_type.size
# Fully Connected
self.fully_net = torch.nn.Sequential(
torch.nn.Linear(c, 64),
torch.nn.BatchNorm1d(64),
torch.nn.ELU(inplace=True),
torch.nn.Linear(64, n_classes),
)