一、 模型如下：
i is 0, m is RFBNet(
(base): ModuleList(
(0): Conv2d(1, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): ReLU(inplace)
(2): Conv2d(24, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(3): ReLU(inplace)
(4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(5): Conv2d(24, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(6): ReLU(inplace)
(7): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(8): ReLU(inplace)
(9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(10): Conv2d(48, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(11): ReLU(inplace)
(12): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(13): ReLU(inplace)
(14): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(15): ReLU(inplace)
(16): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(17): Conv2d(96, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(18): ReLU(inplace)
(19): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(20): ReLU(inplace)
(21): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(22): ReLU(inplace)
)
(Norm): BasicRFB_a(
(branch0): Sequential(
(0): BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
(branch1): Sequential(
(0): BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(2): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
(branch2): Sequential(
(0): BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(2): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
(branch3): Sequential(
(0): BasicConv(
(conv): Conv2d(192, 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(24, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(24, 36, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
(bn): BatchNorm2d(36, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(2): BasicConv(
(conv): Conv2d(36, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(3): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(5, 5), dilation=(5, 5), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
(ConvLinear): BasicConv(
(conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(192, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
(shortcut): BasicConv(
(conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(192, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
(relu): ReLU()
)
(loc): ModuleList(
(0): Conv2d(192, 20, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
)
(conf): ModuleList(
(0): Conv2d(192, 25, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
)
(softmax): Softmax()
)
i is 1, m is ModuleList(
(0): Conv2d(1, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): ReLU(inplace)
(2): Conv2d(24, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(3): ReLU(inplace)
(4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(5): Conv2d(24, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(6): ReLU(inplace)
(7): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(8): ReLU(inplace)
(9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(10): Conv2d(48, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(11): ReLU(inplace)
(12): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(13): ReLU(inplace)
(14): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(15): ReLU(inplace)
(16): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(17): Conv2d(96, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(18): ReLU(inplace)
(19): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(20): ReLU(inplace)
(21): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(22): ReLU(inplace)
)
i is 2, m is Conv2d(1, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 3, m is ReLU(inplace)
i is 4, m is Conv2d(24, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 5, m is ReLU(inplace)
i is 6, m is MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
i is 7, m is Conv2d(24, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 8, m is ReLU(inplace)
i is 9, m is Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 10, m is ReLU(inplace)
i is 11, m is MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
i is 12, m is Conv2d(48, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 13, m is ReLU(inplace)
i is 14, m is Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 15, m is ReLU(inplace)
i is 16, m is Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 17, m is ReLU(inplace)
i is 18, m is MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
i is 19, m is Conv2d(96, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 20, m is ReLU(inplace)
i is 21, m is Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 22, m is ReLU(inplace)
i is 23, m is Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 24, m is ReLU(inplace)
i is 25, m is BasicRFB_a(
(branch0): Sequential(
(0): BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
(branch1): Sequential(
(0): BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(2): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
(branch2): Sequential(
(0): BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(2): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
(branch3): Sequential(
(0): BasicConv(
(conv): Conv2d(192, 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(24, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(24, 36, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
(bn): BatchNorm2d(36, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(2): BasicConv(
(conv): Conv2d(36, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(3): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(5, 5), dilation=(5, 5), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
(ConvLinear): BasicConv(
(conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(192, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
(shortcut): BasicConv(
(conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(192, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
(relu): ReLU()
)
i is 26, m is Sequential(
(0): BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
i is 27, m is BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
i is 28, m is Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
i is 29, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 30, m is ReLU(inplace)
i is 31, m is BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
i is 32, m is Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
i is 33, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 34, m is Sequential(
(0): BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(2): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
i is 35, m is BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
i is 36, m is Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
i is 37, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 38, m is ReLU(inplace)
i is 39, m is BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
i is 40, m is Conv2d(48, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
i is 41, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 42, m is ReLU(inplace)
i is 43, m is BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
i is 44, m is Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
i is 45, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 46, m is Sequential(
(0): BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(2): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
i is 47, m is BasicConv(
(conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
i is 48, m is Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
i is 49, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 50, m is ReLU(inplace)
i is 51, m is BasicConv(
(conv): Conv2d(48, 48, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
i is 52, m is Conv2d(48, 48, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
i is 53, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 54, m is ReLU(inplace)
i is 55, m is BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
i is 56, m is Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(3, 3), dilation=(3, 3), bias=False)
i is 57, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 58, m is Sequential(
(0): BasicConv(
(conv): Conv2d(192, 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(24, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(1): BasicConv(
(conv): Conv2d(24, 36, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
(bn): BatchNorm2d(36, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(2): BasicConv(
(conv): Conv2d(36, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
(3): BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(5, 5), dilation=(5, 5), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
)
i is 59, m is BasicConv(
(conv): Conv2d(192, 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(24, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
i is 60, m is Conv2d(192, 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
i is 61, m is BatchNorm2d(24, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 62, m is ReLU(inplace)
i is 63, m is BasicConv(
(conv): Conv2d(24, 36, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
(bn): BatchNorm2d(36, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
i is 64, m is Conv2d(24, 36, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)
i is 65, m is BatchNorm2d(36, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 66, m is ReLU(inplace)
i is 67, m is BasicConv(
(conv): Conv2d(36, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
(relu): ReLU(inplace)
)
i is 68, m is Conv2d(36, 48, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)
i is 69, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 70, m is ReLU(inplace)
i is 71, m is BasicConv(
(conv): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(5, 5), dilation=(5, 5), bias=False)
(bn): BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
i is 72, m is Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(5, 5), dilation=(5, 5), bias=False)
i is 73, m is BatchNorm2d(48, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 74, m is BasicConv(
(conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(192, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
i is 75, m is Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
i is 76, m is BatchNorm2d(192, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 77, m is BasicConv(
(conv): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn): BatchNorm2d(192, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
)
i is 78, m is Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
i is 79, m is BatchNorm2d(192, eps=1e-05, momentum=0.01, affine=True, track_running_stats=True)
i is 80, m is ReLU()
i is 81, m is ModuleList(
(0): Conv2d(192, 20, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
)
i is 82, m is Conv2d(192, 20, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 83, m is ModuleList(
(0): Conv2d(192, 25, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
)
i is 84, m is Conv2d(192, 25, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
i is 85, m is Softmax()
二、 模型代码
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
from cell_detection.layers import *
import torchvision.transforms as transforms
import torchvision.models as models
import torch.backends.cudnn as cudnn
import os
class BasicConv(nn.Module):
def __init__(self, in_planes, out_planes, kernel_size, stride=1, padding=0, dilation=1, groups=1, relu=True, bn=True, bias=False):
super(BasicConv, self).__init__()
self.out_channels = out_planes
self.conv = nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride, padding=padding, dilation=dilation, groups=groups, bias=bias)
self.bn = nn.BatchNorm2d(out_planes,eps=1e-5, momentum=0.01, affine=True) if bn else None
self.relu = nn.ReLU(inplace=True) if relu else None
def forward(self, x):
x = self.conv(x)
if self.bn is not None:
x = self.bn(x)
if self.relu is not None:
x = self.relu(x)
return x
class BasicRFB(nn.Module):
def __init__(self, in_planes, out_planes, stride=1, scale = 0.1, visual = 1):
super(BasicRFB, self).__init__()
self.scale = scale
self.out_channels = out_planes
inter_planes = in_planes // 8
self.branch0 = nn.Sequential(
BasicConv(in_planes, inter_planes, kernel_size=1, stride=1),
#BasicConv(inter_planes, (inter_planes//2)*3, kernel_size=(1,3), stride=1, padding=(0,1)),
BasicConv(inter_planes, 2*inter_planes, kernel_size=(3,3), stride=stride, padding=(1,1)),
BasicConv(2*inter_planes, 2*inter_planes, kernel_size=3, stride=1, padding=visual+1, dilation=visual+1, relu=False)
)
self.branch1 = nn.Sequential(
BasicConv(in_planes, inter_planes, kernel_size=1, stride=1),
BasicConv(inter_planes, (inter_planes//2)*3, kernel_size=3, stride=1, padding=1),
BasicConv((inter_planes//2)*3, 2*inter_planes, kernel_size=3, stride=stride, padding=1),
BasicConv(2*inter_planes, 2*inter_planes, kernel_size=3, stride=1, padding=2*visual+1, dilation=2*visual+1, relu=False)
)
self.ConvLinear = BasicConv(4*inter_planes, out_planes, kernel_size=1, stride=1, relu=False)
self.shortcut = BasicConv(in_planes, out_planes, kernel_size=1, stride=stride, relu=False)
self.relu = nn.ReLU(inplace=False)
def forward(self,x):
x0 = self.branch0(x)
x1 = self.branch1(x)
out = torch.cat((x0,x1),1)
out = self.ConvLinear(out)
short = self.shortcut(x)
out = out*self.scale + short # 分别经过out和short，再一个线性组合？
out = self.relu(out)
return out
class BasicRFB_a(nn.Module):
def __init__(self, in_planes, out_planes, stride=1, scale = 0.1):
super(BasicRFB_a, self).__init__()
self.scale = scale
self.out_channels = out_planes
inter_planes = in_planes //4
self.branch0 = nn.Sequential(
BasicConv(in_planes, inter_planes, kernel_size=1, stride=1),
BasicConv(inter_planes, inter_planes, kernel_size=3, stride=1, padding=1,relu=False)
)
self.branch1 = nn.Sequential(
BasicConv(in_planes, inter_planes, kernel_size=1, stride=1),
BasicConv(inter_planes, inter_planes, kernel_size=(3,1), stride=1, padding=(1,0)),
BasicConv(inter_planes, inter_planes, kernel_size=3, stride=1, padding=3, dilation=3, relu=False)
)
self.branch2 = nn.Sequential(
BasicConv(in_planes, inter_planes, kernel_size=1, stride=1),
BasicConv(inter_planes, inter_planes, kernel_size=(1,3), stride=stride, padding=(0,1)),
BasicConv(inter_planes, inter_planes, kernel_size=3, stride=1, padding=3, dilation=3, relu=False)
)
'''
self.branch3 = nn.Sequential(
BasicConv(in_planes, inter_planes, kernel_size=1, stride=1),
BasicConv(inter_planes, inter_planes, kernel_size=3, stride=1, padding=1),
BasicConv(inter_planes, inter_planes, kernel_size=3, stride=1, padding=3, dilation=3, relu=False)
)
'''
self.branch3 = nn.Sequential(
BasicConv(in_planes, inter_planes//2, kernel_size=1, stride=1),
BasicConv(inter_planes//2, (inter_planes//4)*3, kernel_size=(1,3), stride=1, padding=(0,1)),
BasicConv((inter_planes//4)*3, inter_planes, kernel_size=(3,1), stride=stride, padding=(1,0)),
BasicConv(inter_planes, inter_planes, kernel_size=3, stride=1, padding=5, dilation=5, relu=False)
)
self.ConvLinear = BasicConv(4*inter_planes, out_planes, kernel_size=1, stride=1, relu=False)
self.shortcut = BasicConv(in_planes, out_planes, kernel_size=1, stride=stride, relu=False)
self.relu = nn.ReLU(inplace=False)
def forward(self, x):
x0 = self.branch0(x)
x1 = self.branch1(x)
x2 = self.branch2(x)
x3 = self.branch3(x)
out = torch.cat((x0,x1,x2,x3),1)
out = self.ConvLinear(out)
short = self.shortcut(x)
out = out*self.scale + short
out = self.relu(out)
return out
class RFBNet(nn.Module):
"""RFB Net for object detection
The network is based on the SSD architecture.
Each multibox layer branches into
1) conv2d for class conf scores
2) conv2d for localization predictions
3) associated priorbox layer to produce default bounding
boxes specific to the layer's feature map size.
See: https://arxiv.org/pdf/1711.07767.pdf for more details on RFB Net.
Args:
phase: (string) Can be "test" or "train"
base: VGG16 layers for input, size of either 300 or 512
extras: extra layers that feed to multibox loc and conf layers
head: "multibox head" consists of loc and conf conv layers
"""
def __init__(self, phase, size, base, extras, head, num_classes):
super(RFBNet, self).__init__()
self.phase = phase
self.num_classes = num_classes
self.size = size
if size == 300:
self.indicator = 3
elif size == 512:
self.indicator = 5
else:
print("Error: Sorry only SSD300 and SSD512 are supported!")
return
# vgg network
self.base = nn.ModuleList(base) # base,extras, head对应了multibox
# conv_4
self.Norm = BasicRFB_a(192, 192, stride = 1, scale=1.0)
# self.extras = nn.ModuleList(extras[0:2]) # 只用conv4_3这一层
self.loc = nn.ModuleList(head[0][0:1])
self.conf = nn.ModuleList(head[1][0:1])
if self.phase == 'test':
self.softmax = nn.Softmax()
def forward(self, x):
"""Applies network layers and ops on input image(s) x.
Args:
x: input image or batch of images. Shape: [batch,3*batch,300,300].
Return:
Depending on phase:
test:
list of concat outputs from:
1: softmax layers, Shape: [batch*num_priors,num_classes]
2: localization layers, Shape: [batch,num_priors*4]
3: priorbox layers, Shape: [2,num_priors*4]
train:
list of concat outputs from:
1: confidence layers, Shape: [batch*num_priors,num_classes]
2: localization layers, Shape: [batch,num_priors*4]
3: priorbox layers, Shape: [2,num_priors*4]
"""
sources = list()
loc = list()
conf = list()
# apply vgg up to conv4_3 relu
for k in range(23):
x = self.base[k](x)
s = self.Norm(x) # BasicRFB_a这个分支
sources.append(s)
# apply multibox head to source layers
for (x, l, c) in zip(sources, self.loc, self.conf):
loc.append(l(x).permute(0, 2, 3, 1).contiguous()) # permute调换维度，channel放到最后
conf.append(c(x).permute(0, 2, 3, 1).contiguous())
#print([o.size() for o in loc])
feature_map_size = [(o.size()[1], o.size()[2]) for o in conf]
loc = torch.cat([o.view(o.size(0), -1) for o in loc], 1) # 将特征直接拉直
conf = torch.cat([o.view(o.size(0), -1) for o in conf], 1)
if self.phase == "test":
output = (
loc.view(loc.size(0), -1, 4), # loc preds
self.softmax(conf.view(-1, self.num_classes)), # conf preds
)
else:
output = (
loc.view(loc.size(0), -1, 4),
conf.view(conf.size(0), -1, self.num_classes),
)
return output, feature_map_size
def load_weights(self, base_file):
other, ext = os.path.splitext(base_file)
if ext == '.pkl' or '.pth':
print('Loading weights into state dict...')
self.load_state_dict(torch.load(base_file))
print('Finished!')
else:
print('Sorry only .pth and .pkl files supported.')
# This function is derived from torchvision VGG make_layers()
# https://github.com/pytorch/vision/blob/master/torchvision/models/vgg.py
def vgg(cfg, i, batch_norm=False):
layers = []
in_channels = i
for v in cfg:
if v == 'M':
layers += [nn.MaxPool2d(kernel_size=2, stride=2)]
elif v == 'C':
layers += [nn.MaxPool2d(kernel_size=2, stride=2)] #pooling的时候使用ceil还是floor
else:
conv2d = nn.Conv2d(in_channels, v, kernel_size=3, padding=1)
if batch_norm:
layers += [conv2d, nn.BatchNorm2d(v), nn.ReLU(inplace=True)]
else:
layers += [conv2d, nn.ReLU(inplace=True)]
in_channels = v
# 这里开始不同
pool5 = nn.MaxPool2d(kernel_size=3, stride=1, padding=1)
conv6 = nn.Conv2d(192, 384, kernel_size=3, padding=6, dilation=6) #dilation 6卷积
conv7 = nn.Conv2d(384, 384, kernel_size=1)
layers += [pool5, conv6,
nn.ReLU(inplace=True), conv7, nn.ReLU(inplace=True)]
return layers
base = {
'512': [24, 24, 'M', 48, 48, 'M', 96, 96, 96, 'C', 192, 192, 192, 'M',
192, 192, 192],
}
def add_extras(size, cfg, i, batch_norm=False):
# Extra layers added to VGG for feature scaling
layers = []
in_channels = i
flag = False
for k, v in enumerate(cfg):
if in_channels != 'S':
if v == 'S':
if in_channels == 96 and size == 512:
layers += [BasicRFB(in_channels, cfg[k+1], stride=2, scale = 1.0, visual=1)]
else:
layers += [BasicRFB(in_channels, cfg[k+1], stride=2, scale = 1.0, visual=2)]
else:
layers += [BasicRFB(in_channels, v, scale = 1.0, visual=2)]
in_channels = v
if size == 512:
layers += [BasicConv(96,48,kernel_size=1,stride=1)]
layers += [BasicConv(48,96,kernel_size=4,stride=1,padding=1)]
elif size ==300:
layers += [BasicConv(256,128,kernel_size=1,stride=1)]
layers += [BasicConv(128,256,kernel_size=3,stride=1)]
layers += [BasicConv(256,128,kernel_size=1,stride=1)]
layers += [BasicConv(128,256,kernel_size=3,stride=1)]
else:
print("Error: Sorry only RFBNet300 and RFBNet512 are supported!")
return
return layers
extras = {
'512': [384, 'S', 192, 'S', 96, 'S', 96,'S',96],
}
def multibox(size, vgg, extra_layers, cfg, num_classes):
"""
主要是用来生成loc_layers和conf_layers
:param size: image size
:param vgg: 传入的vgg网络，list类型，包含各个层
:param extra_layers: list类型
:param cfg: [6,6,6,6,4,4]list, mbox中定义的
:param num_classes:
:return:
"""
loc_layers = []
conf_layers = []
vgg_source = [-2]
for k, v in enumerate(vgg_source):
if k == 0:
loc_layers += [nn.Conv2d(192,
cfg[k] * 4, kernel_size=3, padding=1)]
conf_layers +=[nn.Conv2d(192,
cfg[k] * num_classes, kernel_size=3, padding=1)]
else:
loc_layers += [nn.Conv2d(vgg[v].out_channels,
cfg[k] * 4, kernel_size=3, padding=1)]
conf_layers += [nn.Conv2d(vgg[v].out_channels,
cfg[k] * num_classes, kernel_size=3, padding=1)]
i = 1
indicator = 0
if size == 300:
indicator = 3
elif size == 512:
indicator = 5
else:
print("Error: Sorry only RFBNet300 and RFBNet512 are supported!")
return
for k, v in enumerate(extra_layers):
if k < indicator or k%2== 0:
loc_layers += [nn.Conv2d(v.out_channels, cfg[i]
* 4, kernel_size=3, padding=1)]
conf_layers += [nn.Conv2d(v.out_channels, cfg[i]
* num_classes, kernel_size=3, padding=1)]
i +=1
return vgg, extra_layers, (loc_layers, conf_layers)
mbox = {
'512': [5, 5, 5, 5, 5, 4, 4], # number of boxes per feature map location
}
def build_net(phase, size=300, num_classes=21):
if phase != "test" and phase != "train":
print("Error: Phase not recognized")
return
if size != 300 and size != 512:
print("Error: Sorry only RFBNet300 and RFBNet512 are supported!")
return
return RFBNet(phase, size, *multibox(size, vgg(base[str(size)], 1),
add_extras(size, extras[str(size)], 384),
mbox[str(size)], num_classes), num_classes)二、计算模型参数占用的显存以及中间变量占用的显存
def modelsize(model, input, type_size=4):
L = list()
L_tmp = list()
for p in model.parameters():
#print("p is: ", p)
L.append(p.size())
L_tmp.append(np.prod(L))
L = list()
para = sum(L_tmp)
print('model {} : params: {:4f}M'.format(model._get_name(), para * type_size / 1000 / 1000))
input_ = input.clone()
input_.requires_grad_(requires_grad=False)
print("model.modules is: ", model.modules())
print("model.modules length is: ", len(list(model.modules())))
mods = list(model.modules())
out_sizes = []
print("***************************************")
for i in range(0, len(mods)):
m = mods[i]
#print("i is {0}, m is {1}".format(i, m))
print("***************************************")
for i in range(2, 24):
m = mods[i]
#print("i is {0}, m is {1}".format(i, m))
if isinstance(m, nn.ReLU):
if m.inplace:
continue
out = m(input_)
out_sizes.append(np.array(out.size()))
input_ = out
input_0 = input_.clone()
for i in range(28, 34):
if i == 31:
continue
m = mods[i]
#print("i is {0}, m is {1}".format(i, m))
#print("input_shape size is: ", input_0.shape)
if isinstance(m, nn.ReLU):
if m.inplace:
continue
out_0 = m(input_0)
out_sizes.append(np.array(out_0.size()))
input_0 = out_0
input_1 = input_.clone()
for i in range(36, 46):
if i == 39 or i == 43 :
continue
m = mods[i]
#print("i is {0}, m is {1}".format(i, m))
if isinstance(m, nn.ReLU):
if m.inplace:
continue
out_1 = m(input_1)
out_sizes.append(np.array(out_1.size()))
input_1 = out_1
input_2 = input_.clone()
for i in range(48, 58):
if i == 51 or i == 55 :
continue
m = mods[i]
#print("i is {0}, m is {1}".format(i, m))
if isinstance(m, nn.ReLU):
if m.inplace:
continue
out_2 = m(input_2)
out_sizes.append(np.array(out_2.size()))
input_2 = out_2
input_3 = input_.clone()
for i in range(60, 74):
if i == 63 or i == 67 or i == 71:
continue
m = mods[i]
print("i is {0}, m is {1}".format(i, m))
if isinstance(m, nn.ReLU):
if m.inplace:
continue
out_3 = m(input_3)
out_sizes.append(np.array(out_3.size()))
input_3 = out_3
input_RFB = torch.cat((input_0, input_1, input_2, input_3),1)
for i in range(75, 81):
if i == 77:
continue
m = mods[i]
print("i is {0}, m is {1}".format(i, m))
if isinstance(m, nn.ReLU):
if m.inplace:
continue
out_RFB = m(input_RFB)
out_sizes.append(np.array(out_RFB.size()))
input_RFB = out_RFB
total_nums = 0
for i in range(len(out_sizes)):
s = out_sizes[i]
nums = np.prod(np.array(s))
total_nums += nums
print('Model {} : intermedite variables: {:3f} M (without backward)'.format(model._get_name(), total_nums * type_size / 1000 / 1000))参考博客：