def forward(self, x):
residual = x
# Point-wise expansion
x = self.conv_pw(x)
x = self.bn1(x)
x = self.act1(x)
# Depth-wise convolution
x = self.conv_dw(x)
x = self.bn2(x)
x = self.act2(x)
# Squeeze-and-excitation
if self.se is not None:
x = self.se(x)
# Point-wise linear projection
x = self.conv_pwl(x)
x = self.bn3(x)
if self.has_residual:
if self.downsample is not None:
residual = self.downsample(residual)
if self.drop_path_rate > 0.:
x = drop_path(x, self.drop_path_rate, self.training)
x += residual
return x
def forward(self, x): if self.conv is None or isinstance(self.conv, ZeroLayer): res = x elif self.shortcut is None or isinstance(self.shortcut, ZeroLayer): res = self.conv(x) else: shortcut = self.shortcut(x) x = self.conv(x) if self.drop_path_rate > 0.: x = drop_path(x, self.drop_path_rate, self.training) res = x + shortcut return res
def forward(self, x):
outputs = []
for level in range(self.config.num_levels):
x_level = x[level]
for i in range(self.config.box_class_repeats):
x_level_ident = x_level
x_level = self.conv_rep[i](x_level)
x_level = self.bn_rep[i][level](x_level)
x_level = self.act(x_level)
if i > 0 and self.config.fpn_drop_path_rate:
x_level = drop_path(x_level, self.config.fpn_drop_path_rate, self.training)
x_level += x_level_ident
outputs.append(self.predict(x_level))
return outputs
def forward(self, x):
if self.mobile_inverted_conv is None or isinstance(self.mobile_inverted_conv, ZeroLayer):
res = x
elif self.shortcut is None or isinstance(self.shortcut, ZeroLayer):
res = self.mobile_inverted_conv(x)
else:
# res = self.mobile_inverted_conv(x) + self.shortcut(x)
res = self.mobile_inverted_conv(x)
if self.drop_connect_rate > 0.:
res = drop_path(res, drop_prob=self.drop_connect_rate, training=self.training)
res += self.shortcut(x)
return res
def forward(self, x):
return drop_path(x, self.drop_prob, self.training)
