import torch
from torch import nn
from torch.nn import functional as F
from EfficientNet_2d.utils import (
round_filters,
round_repeats,
drop_connect,
get_same_padding_conv2d,
get_model_params,
efficientnet_params,
load_pretrained_weights,
Swish,
MemoryEfficientSwish,
)
class MBConvBlock(nn.Module):
"""
Mobile Inverted Residual Bottleneck Block
Args:
block_args (namedtuple): BlockArgs, see above
global_params (namedtuple): GlobalParam, see above
Attributes:
has_se (bool): Whether the block contains a Squeeze and Excitation layer.
"""
def __init__(self, block_args, global_params):
super().__init__()
self._block_args = block_args
self._bn_mom = 1 - global_params.batch_norm_momentum
self._bn_eps = global_params.batch_norm_epsilon
self.has_se = (self._block_args.se_ratio is not None) and (0 < self._block_args.se_ratio <= 1)
self.id_skip = block_args.id_skip # skip connection and drop connect
# Get static or dynamic convolution depending on image size
Conv2d = get_same_padding_conv2d(image_size=global_params.image_size)
# Expansion phase
inp = self._block_args.input_filters # number of input channels
oup = self._block_args.input_filters * self._block_args.expand_ratio # number of output channels
if self._block_args.expand_ratio != 1:
self._expand_conv = Conv2d(in_channels=inp, out_channels=oup, kernel_size=1, bias=False)
self._bn0 = nn.BatchNorm2d(num_features=oup, momentum=self._bn_mom, eps=self._bn_eps)
# Depthwise convolution phase
k = self._block_args.kernel_size
s = self._block_args.stride
self._depthwise_conv = Conv2d(
in_channels=oup, out_channels=oup, groups=oup, # groups makes it depthwise
kernel_size=k, stride=s, bias=False)
self._bn1 = nn.BatchNorm2d(num_features=oup, momentum=self._bn_mom, eps=self._bn_eps)
# Squeeze and Excitation layer, if desired
if self.has_se:
num_squeezed_channels = max(1, int(self._block_args.input_filters * self._block_args.se_ratio))
self._se_reduce = Conv2d(in_channels=oup, out_channels=num_squeezed_channels, kernel_size=1)
self._se_expand = Conv2d(in_channels=num_squeezed_channels, out_channels=oup, kernel_size=1)
# Output phase
final_oup = self._block_args.output_filters
self._project_conv = Conv2d(in_channels=oup, out_channels=final_oup, kernel_size=1, bias=False)
self._bn2 = nn.BatchNorm2d(num_features=final_oup, momentum=self._bn_mom, eps=self._bn_eps)
self._swish = MemoryEfficientSwish()
def forward(self, inputs, drop_connect_rate=None):
"""
:param inputs: input tensor
:param drop_connect_rate: drop connect rate (float, between 0 and 1)
:return: output of block
"""
# Expansion and Depthwise Convolution
x = inputs
if self._block_args.expand_ratio != 1:
x = self._swish(self._bn0(self._expand_conv(inputs)))
x = self._swish(self._bn1(self._depthwise_conv(x)))
# Squeeze and Excitation
if self.has_se:
x_squeezed = F.adaptive_avg_pool2d(x, 1)
x_squeezed = self._se_expand(self._swish(self._se_reduce(x_squeezed)))
x = torch.sigmoid(x_squeezed) * x
x = self._bn2(self._project_conv(x))
# Skip connection and drop connect
input_filters, output_filters = self._block_args.input_filters, self._block_args.output_filters
if self.id_skip and self._block_args.stride == 1 and input_filters == output_filters:
if drop_connect_rate:
x = drop_connect(x, p=drop_connect_rate, training=self.training)
x = x + inputs # skip connection
return x
def set_swish(self, memory_efficient=True):
"""Sets swish function as memory efficient (for training) or standard (for export)"""
self._swish = MemoryEfficientSwish() if memory_efficient else Swish()
class EfficientNet(nn.Module):
"""
An EfficientNet model. Most easily loaded with the .from_name or .from_pretrained methods
Args:
blocks_args (list): A list of BlockArgs to construct blocks
global_params (namedtuple): A set of GlobalParams shared between blocks
Example:
model = EfficientNet.from_pretrained('efficientnet-b0')
"""
def __init__(self, blocks_args=None, global_params=None):
super().__init__()
assert isinstance(blocks_args, list), 'blocks_args should be a list'
assert len(blocks_args) > 0, 'block args must be greater than 0'
self._global_params = global_params
self._blocks_args = blocks_args
# Get static or dynamic convolution depending on image size
Conv2d = get_same_padding_conv2d(image_size=global_params.image_size)
# Batch norm parameters
bn_mom = 1 - self._global_params.batch_norm_momentum
bn_eps = self._global_params.batch_norm_epsilon
# Stem
in_channels = 3 # rgb
out_channels = round_filters(32, self._global_params) # number of output channels
self._conv_stem = Conv2d(in_channels, out_channels, kernel_size=3, stride=2, bias=False)
self._bn0 = nn.BatchNorm2d(num_features=out_channels, momentum=bn_mom, eps=bn_eps)
# Build blocks
self._blocks = nn.ModuleList([])
for block_args in self._blocks_args:
# Update block input and output filters based on depth multiplier.
block_args = block_args._replace(
input_filters=round_filters(block_args.input_filters, self._global_params),
output_filters=round_filters(block_args.output_filters, self._global_params),
num_repeat=round_repeats(block_args.num_repeat, self._global_params)
)
# The first block needs to take care of stride and filter size increase.
self._blocks.append(MBConvBlock(block_args, self._global_params))
if block_args.num_repeat > 1:
block_args = block_args._replace(input_filters=block_args.output_filters, stride=1)
for _ in range(block_args.num_repeat - 1):
self._blocks.append(MBConvBlock(block_args, self._global_params))
# Head
in_channels = block_args.output_filters # output of final block
out_channels = round_filters(1280, self._global_params)
self._conv_head = Conv2d(in_channels, out_channels, kernel_size=1, bias=False)
self._bn1 = nn.BatchNorm2d(num_features=out_channels, momentum=bn_mom, eps=bn_eps)
# Final linear layer
self._avg_pooling = nn.AdaptiveAvgPool2d(1)
self._dropout = nn.Dropout(self._global_params.dropout_rate)
self._fc = nn.Linear(out_channels, self._global_params.num_classes)
self._swish = MemoryEfficientSwish()
def set_swish(self, memory_efficient=True):
"""Sets swish function as memory efficient (for training) or standard (for export)"""
self._swish = MemoryEfficientSwish() if memory_efficient else Swish()
for block in self._blocks:
block.set_swish(memory_efficient)
def extract_features(self, inputs):
""" Returns output of the final convolution layer """
# Stem
x = self._swish(self._bn0(self._conv_stem(inputs)))
# Blocks
for idx, block in enumerate(self._blocks):
drop_connect_rate = self._global_params.drop_connect_rate
if drop_connect_rate:
drop_connect_rate *= float(idx) / len(self._blocks)
x = block(x, drop_connect_rate=drop_connect_rate)
# Head
x = self._swish(self._bn1(self._conv_head(x)))
return x
def forward(self, inputs):
""" Calls extract_features to extract features, applies final linear layer, and returns logits. """
bs = inputs.size(0)
# Convolution layers
x = self.extract_features(inputs)
# Pooling and final linear layer
x = self._avg_pooling(x)
x = x.view(bs, -1)
x = self._dropout(x)
x = self._fc(x)
return x
@classmethod
def from_name(cls, model_name, override_params=None):
cls._check_model_name_is_valid(model_name)
blocks_args, global_params = get_model_params(model_name, override_params)
return cls(blocks_args, global_params)
@classmethod
def from_pretrained(cls, model_name, advprop=False, num_classes=1000, in_channels=3):
model = cls.from_name(model_name, override_params={'num_classes': num_classes})
load_pretrained_weights(model, model_name, load_fc=(num_classes == 1000), advprop=advprop)
if in_channels != 3:
Conv2d = get_same_padding_conv2d(image_size = model._global_params.image_size)
out_channels = round_filters(32, model._global_params)
model._conv_stem = Conv2d(in_channels, out_channels, kernel_size=3, stride=2, bias=False)
return model
@classmethod
def get_image_size(cls, model_name):
cls._check_model_name_is_valid(model_name)
_, _, res, _ = efficientnet_params(model_name)
return res
@classmethod
def _check_model_name_is_valid(cls, model_name):
""" Validates model name. """
valid_models = ['efficientnet-b'+str(i) for i in range(9)]
if model_name not in valid_models:
raise ValueError('model_name should be one of: ' + ', '.join(valid_models))
# get pretrained EfficientNet for k-classes classification
def get_pretrained_EfficientNet(num_classes):
model = EfficientNet.from_pretrained('efficientnet-b0')
fc_features = model._fc.in_features
model._fc = nn.Linear(fc_features, num_classes)
return model
class DAR_Effi(nn.Module):
def __init__(self, blocks_args=None, global_params=None, in_channels=3, att_start=11):
super(DAR_Effi, self).__init__()
assert isinstance(blocks_args, list), 'blocks_args should be a list'
assert len(blocks_args) > 0, 'block args must be greater than 0'
self._global_params = global_params
self._blocks_args = blocks_args
self.att_start = att_start # for CA-module and NA-module
# Get static or dynamic convolution depending on image size
Conv2d = get_same_padding_conv2d(image_size=global_params.image_size)
# Batch norm parameters
bn_mom = 1 - self._global_params.batch_norm_momentum
bn_eps = self._global_params.batch_norm_epsilon
# Stem
out_channels = round_filters(32, self._global_params) # number of output channels
self._conv_stem = Conv2d(in_channels, out_channels, kernel_size=3, stride=2, bias=False)
self._bn0 = nn.BatchNorm2d(num_features=out_channels, momentum=bn_mom, eps=bn_eps)
self._conv_stem_cf = Conv2d(in_channels, out_channels, kernel_size=3, stride=2, bias=False)
self._bn0_cf = nn.BatchNorm2d(num_features=out_channels, momentum=bn_mom, eps=bn_eps)
self._conv_stem_lr = Conv2d(in_channels, out_channels, kernel_size=3, stride=2, bias=False)
self._bn0_lr = nn.BatchNorm2d(num_features=out_channels, momentum=bn_mom, eps=bn_eps)
# Build blocks of Prd-Net
self._blocks = nn.ModuleList([])
for block_args in self._blocks_args:
# Update block input and output filters based on depth multiplier.
block_args = block_args._replace(
input_filters=round_filters(block_args.input_filters, self._global_params),
output_filters=round_filters(block_args.output_filters, self._global_params),
num_repeat=round_repeats(block_args.num_repeat, self._global_params)
)
# The first block needs to take care of stride and filter size increase.
self._blocks.append(MBConvBlock(block_args, self._global_params))
if block_args.num_repeat > 1:
block_args = block_args._replace(input_filters=block_args.output_filters, stride=1)
for _ in range(block_args.num_repeat - 1):
self._blocks.append(MBConvBlock(block_args, self._global_params))
# Build blocks of CF-Net
self._blocks_cf = nn.ModuleList([])
for block_args in self._blocks_args:
# Update block input and output filters based on depth multiplier.
block_args = block_args._replace(
input_filters=round_filters(block_args.input_filters, self._global_params),
output_filters=round_filters(block_args.output_filters, self._global_params),
num_repeat=round_repeats(block_args.num_repeat, self._global_params)
)
# The first block needs to take care of stride and filter size increase.
self._blocks_cf.append(MBConvBlock(block_args, self._global_params))
if block_args.num_repeat > 1:
block_args = block_args._replace(input_filters=block_args.output_filters, stride=1)
for _ in range(block_args.num_repeat - 1):
self._blocks_cf.append(MBConvBlock(block_args, self._global_params))
# Build blocks of LR-Net
self._blocks_lr = nn.ModuleList([])
for block_args in self._blocks_args:
# Update block input and output filters based on depth multiplier.
block_args = block_args._replace(
input_filters=round_filters(block_args.input_filters, self._global_params),
output_filters=round_filters(block_args.output_filters, self._global_params),
num_repeat=round_repeats(block_args.num_repeat, self._global_params)
)
# The first block needs to take care of stride and filter size increase.
self._blocks_lr.append(MBConvBlock(block_args, self._global_params))
if block_args.num_repeat > 1:
block_args = block_args._replace(input_filters=block_args.output_filters, stride=1)
for _ in range(block_args.num_repeat - 1):
self._blocks_lr.append(MBConvBlock(block_args, self._global_params))
# Head
in_channels = block_args.output_filters # output of final block
out_channels = round_filters(1280, self._global_params)
self._conv_head = Conv2d(in_channels, out_channels, kernel_size=1, bias=False)
self._bn1 = nn.BatchNorm2d(num_features=out_channels, momentum=bn_mom, eps=bn_eps)
self._conv_head_cf = Conv2d(in_channels, out_channels, kernel_size=1, bias=False)
self._bn1_cf = nn.BatchNorm2d(num_features=out_channels, momentum=bn_mom, eps=bn_eps)
self._conv_head_lr = Conv2d(in_channels, out_channels, kernel_size=1, bias=False)
self._bn1_lr = nn.BatchNorm2d(num_features=out_channels, momentum=bn_mom, eps=bn_eps)
# Final linear layer
self._avg_pooling = nn.AdaptiveAvgPool2d(1)
self._dropout = nn.Dropout(self._global_params.dropout_rate)
self._fc = nn.Linear(out_channels, self._global_params.num_classes)
self._swish = MemoryEfficientSwish()
self._avg_pooling_cf = nn.AdaptiveAvgPool2d(1)
self._dropout_cf = nn.Dropout(self._global_params.dropout_rate)
self._fc_cf = nn.Linear(out_channels, self._global_params.num_classes)
self._swish_cf = MemoryEfficientSwish()
self._avg_pooling_lr = nn.AdaptiveAvgPool2d(1)
self._dropout_lr = nn.Dropout(self._global_params.dropout_rate)
self._fc_lr = nn.Linear(out_channels, self._global_params.num_classes)
self._swish_lr = MemoryEfficientSwish()
def set_swish(self, memory_efficient=True):
"""Sets swish function as memory efficient (for training) or standard (for export)"""
self._swish = MemoryEfficientSwish() if memory_efficient else Swish()
for block in self._blocks:
block.set_swish(memory_efficient)
self._swish_cf = MemoryEfficientSwish() if memory_efficient else Swish()
for block_cf in self._blocks_cf:
block_cf.set_swish(memory_efficient)
self._swish_lr = MemoryEfficientSwish() if memory_efficient else Swish()
for block_lr in self._blocks_lr:
block_lr.set_swish(memory_efficient)
def attention(self, f_prd, f_cf, f_lr):
w_cf = 1 - torch.sigmoid(f_cf)
add_cf = w_cf * f_prd
w_lr = 1 - abs(torch.sigmoid(f_prd)-torch.sigmoid(f_lr))
add_lr = w_lr * f_prd
f_prd = f_prd + add_cf + add_lr
return f_prd
def extract_features(self, inputs):
""" Returns output of the final convolution layer """
# Stem
x = self._swish(self._bn0(self._conv_stem(inputs)))
x_cf = self._swish_cf(self._bn0_cf(self._conv_stem_cf(inputs)))
x_lr = self._swish_lr(self._bn0_lr(self._conv_stem_lr(inputs)))
# Blocks
for idx, block in enumerate(self._blocks):
block_cf = self._blocks_cf[idx]
block_lr = self._blocks_lr[idx]
drop_connect_rate = self._global_params.drop_connect_rate
if drop_connect_rate:
drop_connect_rate *= float(idx) / len(self._blocks)
x = block(x, drop_connect_rate=drop_connect_rate)
x_cf = block_cf(x_cf, drop_connect_rate=drop_connect_rate)
x_lr = block_lr(x_lr, drop_connect_rate=drop_connect_rate)
if idx >= self.att_start:
x = self.attention(x, x_cf, x_lr)
# Head
x = self._swish(self._bn1(self._conv_head(x)))
x_cf = self._swish_cf(self._bn1_cf(self._conv_head_cf(x_cf)))
x_lr = self._swish_lr(self._bn1_lr(self._conv_head_lr(x_lr)))
return x, x_cf, x_lr
def forward(self, inputs):
bs = inputs.size(0)
# Convolution layers
x, x_cf, x_lr = self.extract_features(inputs)
# Pooling and final linear layer
x = self._avg_pooling(x)
x = x.view(bs, -1)
x = self._dropout(x)
x = self._fc(x)
x_cf = self._avg_pooling_cf(x_cf)
x_cf = x_cf.view(bs, -1)
x_cf = self._dropout_cf(x_cf)
x_cf = self._fc_cf(x_cf)
x_lr = self._avg_pooling_lr(x_lr)
x_lr = x_lr.view(bs, -1)
x_lr = self._dropout_lr(x_lr)
x_lr = self._fc_lr(x_lr)
return x, x_cf, x_lr
@classmethod
def from_name(cls, model_name, override_params=None, in_channels=3, att_start=11):
cls._check_model_name_is_valid(model_name)
blocks_args, global_params = get_model_params(model_name, override_params)
return cls(blocks_args, global_params, in_channels, att_start)
@classmethod
def get_image_size(cls, model_name):
cls._check_model_name_is_valid(model_name)
_, _, res, _ = efficientnet_params(model_name)
return res
@classmethod
def _check_model_name_is_valid(cls, model_name):
""" Validates model name. """
valid_models = ['efficientnet-b'+str(i) for i in range(9)]
if model_name not in valid_models:
raise ValueError('model_name should be one of: ' + ', '.join(valid_models))
def get_pretrained_DAR(prd_params, cf_params, lr_params, num_classes):
dar_model = DAR_Effi.from_name('efficientnet-b0')
fc_features = dar_model._fc.in_features
dar_model._fc = nn.Linear(fc_features, num_classes)
dar_model._fc_cf = nn.Linear(fc_features, num_classes)
dar_model._fc_lr = nn.Linear(fc_features, num_classes)
dar_params = dar_model.state_dict()
for k, v in prd_params.items():
index_point = k.find('.')
k_apart = k[0:index_point]
k_bpart = k[index_point:len(k)]
k_cf = k_apart + '_cf' + k_bpart
k_lr = k_apart + '_lr' + k_bpart
dar_params[k] = prd_params[k]
dar_params[k_cf] = cf_params[k]
dar_params[k_lr] = lr_params[k]
dar_model.load_state_dict(dar_params)
return dar_model
Datasets
Models
