An important question !!!

[y78h11b09]

I review your code , and have a fusion that is different from the orginal paper, and furthermore yet exist some different between
self.p4_down_channel = nn.Sequential(
Conv2dStaticSamePadding(conv_channels[1], num_channels, 1),
nn.BatchNorm2d(num_channels, momentum=0.01, eps=1e-3),
)
self.p5_down_channel= nn.Sequential(
Conv2dStaticSamePadding(conv_channels[2], num_channels, 1),
nn.BatchNorm2d(num_channels, momentum=0.01, eps=1e-3),
)
and
self.p4_down_channel_2 = nn.Sequential(
Conv2dStaticSamePadding(conv_channels[1], num_channels, 1),
nn.BatchNorm2d(num_channels, momentum=0.01, eps=1e-3),
)
self.p5_down_channel_2 = nn.Sequential(
Conv2dStaticSamePadding(conv_channels[2], num_channels, 1),
nn.BatchNorm2d(num_channels, momentum=0.01, eps=1e-3),
)

why not use self.p4_down_channel and self.p5_down_channel in where is used self.p4_down_channel_2 and self.p5_down_channel_2 ????

[zylo117]

duplicated.
#412