OpenPose MobileNet V1 作爲Backbone的模型結構（1）
MobileNet V1 的改造
flyfish
論文地址
論文中提到的源碼地址
深度可分離卷積是如何減少計算量的
OpenPose MobileNet V1 作爲Backbone的模型結構（1） MobileNet V1 的改造
OpenPose MobileNet V1 作爲Backbone的模型結構（2） Initial Stage和Refinement Stage的優化
OpenPose MobileNet V1 作爲Backbone的模型結構（3） Branch的合併
原來的MobileNet v1，經過改造之後，名字改爲Dilated MobileNet v1
VGG作爲Backbone的模型結構，OpenPose 是直接取一些層不做更改;MobileNet V1 作爲Backbone的模型結構是，OpenPose做少許更改
看下面原始的MobileNetV1，再做說明是如何更改的
MobileNet_Original(
(model): Sequential(
(0): Sequential(
(0): Conv2d(3, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
(1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(32, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(2): Sequential(
(0): Conv2d(64, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=64, bias=False)
(1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(3): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(4): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=128, bias=False)
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(5): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256, bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(6): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=256, bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(7): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(8): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(9): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(10): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(11): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(12): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(13): Sequential(
(0): Conv2d(1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1024, bias=False)
(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(14): AvgPool2d(kernel_size=7, stride=7, padding=0)
)
(fc): Linear(in_features=1024, out_features=1000, bias=True)
)
更改1
去除，12,13,14平均池化層和最後的15全連接層
(12): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(13): Sequential(
(0): Conv2d(1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1024, bias=False)
(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(1024, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(14): AvgPool2d(kernel_size=7, stride=7, padding=0)
)
(fc): Linear(in_features=1024, out_features=1000, bias=True)
再看6和7
(6): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=256, bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(7): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(6)(0)的stride由stride=(2, 2)更改爲 stride=(1, 1)
更改之後是
(6): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256, bias=False)
(7)(0)使用了空洞卷積（dilated convolution）https://arxiv.org/abs/1511.07122
(7): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), groups=512, bias=False)
原始論文是這樣說的
To save spatial resolution and reuse backbone weights we use dilated convolution . Stride of conv4_2/dw layer was removed and dilation parameter value was set to 2 for succeeding conv5_1/dw layer to preserve receptive field. So we use all layers till conv5_5 block.
Addition of conv5_6 block improves the accuracy, but at cost of performance. We also tried more lightweight backbone MobileNet v2, however it did not show good result, see Table 4.
爲了節省空間分辨率和重複使用backbone權重，我們使用膨脹卷積。去除Conv4_2/dw層的步幅，並將隨後的Conv5_1/dw層的dilation參數值設置爲2，以保留感受野。因此，我們使用所有的層，直到conv5_5 block。加入Conv5_6 block提高了精度，但以犧牲性能爲代價。我們也嘗試了更輕量級的主幹MobileNet v2，但是沒有顯示出很好的效果，請參見表4。
Conv4_2/dw對應(6)(0)
Conv5_1/dw(7)(0)
表達的意思就是
conv( 3, 32, stride=2, bias=False),
conv_dw( 32, 64),
conv_dw( 64, 128, stride=2),
conv_dw(128, 128),
conv_dw(128, 256, stride=2),
conv_dw(256, 256),
conv_dw(256, 512), # conv4_2
conv_dw(512, 512, dilation=2, padding=2),
conv_dw(512, 512),
conv_dw(512, 512),
conv_dw(512, 512),
conv_dw(512, 512) # conv5_5
詳細些
conv( 3, 32, stride=2, bias=False), # conv+BN+ReLU
conv_dw( 32, 64), # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw( 64, 128, stride=2), # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw(128, 128), # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw(128, 256, stride=2), # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw(256, 256), # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw(256, 512), # conv4_2 # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw(512, 512, dilation=2, padding=2), # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw(512, 512), # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw(512, 512), # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw(512, 512), # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
conv_dw(512, 512) # conv5_5 # dw_conv(in,in, stride)+BN+ReLU + conv(in,out)+BN+ReLU
VGG的OpenPose作者在VGG後加的
(23): Conv2d(512, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(24): ReLU(inplace=True)
(25): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(26): ReLU(inplace=True)
在MobileNet v1的OpenPose 作者改爲 減少了channel數，從256減到128. 使用3個深度可分離卷積（depthwise separable conv）代替。
(cpm): Cpm(
(align): Sequential(
(0): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(trunk): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(1): ELU(alpha=1.0, inplace=True)
(2): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(3): ELU(alpha=1.0, inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(1): ELU(alpha=1.0, inplace=True)
(2): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(3): ELU(alpha=1.0, inplace=True)
)
(2): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(1): ELU(alpha=1.0, inplace=True)
(2): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(3): ELU(alpha=1.0, inplace=True)
)
)
(conv): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): ReLU(inplace=True)
)
)
Backbone結構更改完了，還有Head的改進即包括Initial stage 和 Refinement Stage
整體結構如下
MobileNetV1_OpenPose(
(model): Sequential(
(0): Sequential(
(0): Conv2d(3, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
(1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(32, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(2): Sequential(
(0): Conv2d(64, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=64, bias=False)
(1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(3): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(4): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=128, bias=False)
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(5): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256, bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(6): Sequential(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256, bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(7): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(8): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(9): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(10): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
(11): Sequential(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
)
)
(cpm): Cpm(
(align): Sequential(
(0): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(trunk): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(1): ELU(alpha=1.0, inplace=True)
(2): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(3): ELU(alpha=1.0, inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(1): ELU(alpha=1.0, inplace=True)
(2): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(3): ELU(alpha=1.0, inplace=True)
)
(2): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
(1): ELU(alpha=1.0, inplace=True)
(2): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(3): ELU(alpha=1.0, inplace=True)
)
)
(conv): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): ReLU(inplace=True)
)
)
(initial_stage): InitialStage(
(trunk): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): ReLU(inplace=True)
)
(2): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): ReLU(inplace=True)
)
)
(heatmaps): Sequential(
(0): Sequential(
(0): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(512, 19, kernel_size=(1, 1), stride=(1, 1))
)
)
(pafs): Sequential(
(0): Sequential(
(0): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(512, 38, kernel_size=(1, 1), stride=(1, 1))
)
)
)
(refinement_stages): ModuleList(
(0): RefinementStage(
(trunk): Sequential(
(0): RefinementStageBlock(
(initial): Sequential(
(0): Conv2d(185, 128, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(trunk): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
)
)
(1): RefinementStageBlock(
(initial): Sequential(
(0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(trunk): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
)
)
(2): RefinementStageBlock(
(initial): Sequential(
(0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(trunk): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
)
)
(3): RefinementStageBlock(
(initial): Sequential(
(0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(trunk): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
)
)
(4): RefinementStageBlock(
(initial): Sequential(
(0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(trunk): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2))
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
)
)
)
(heatmaps): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 19, kernel_size=(1, 1), stride=(1, 1))
)
)
(pafs): Sequential(
(0): Sequential(
(0): Conv2d(128, 128, kernel_size=(1, 1), stride=(1, 1))
(1): ReLU(inplace=True)
)
(1): Sequential(
(0): Conv2d(128, 38, kernel_size=(1, 1), stride=(1, 1))
)
)
)
)
)