yolov5 加入可形变卷积

更新,之前的deformconv2d没加调制参数，应该是dcnv1
现在的才是dcnv2

from torchvision.ops import DeformConv2d

class DCNConv(nn.Module):
    # Standard convolution
    def __init__(self,
                 in_channels,
                 out_channels,
                 kernel_size,
                 stride=1,
                 padding=0,
                 dilation=1, act=True):
    # def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True):  # ch_in, ch_out, kernel, stride, padding, groups
        super().__init__()
        # self.conv1 = nn.Conv2d(c1, c2, 3, 2, 1, groups=g, bias=False)
        deformable_groups = 1
        offset_channels = 18
        self.conv2_offset = nn.Conv2d(in_channels, deformable_groups * offset_channels, kernel_size=3, padding=1)
        self.modulator_conv = nn.Conv2d(in_channels, 
                                     offset_channels//2,
                                     kernel_size=3, 
                                     stride=1,
                                     padding=1, 
                                     bias=True)
        self.conv2 = DeformConv2d(in_channels, out_channels, kernel_size=3, padding=1, bias=False)
        

        self.bn1 = nn.BatchNorm2d(out_channels)

        self.act1 = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
        self.bn2 = nn.BatchNorm2d(out_channels)
        self.act2 = nn.ReLU(inplace=True) if act is True else (act if isinstance(act, nn.Module) else nn.Identity())

    def forward(self, x):
        # print(x.shape)
        # print('-'*50)
        # x = self.act1(self.bn1(self.conv1(x)))
        # print(x.shape)
        offset = self.conv2_offset(x)
        modulator = self.modulator_conv(x)
        x = self.act2(self.bn2(self.conv2(x,offset,modulator)))

        return x

---

修改common.py 文件

from torchvision.ops import DeformConv2d

class DCNConv(nn.Module):
    # Standard convolution
    def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True):  # ch_in, ch_out, kernel, stride, padding, groups
        super().__init__()
        self.conv1 = nn.Conv2d(c1, c2, 3, 2, 1, groups=g, bias=False)
        deformable_groups = 1
        offset_channels = 18
        self.conv2_offset = nn.Conv2d(c2, deformable_groups * offset_channels, kernel_size=3, padding=1)
        self.conv2 = DeformConv2d(c2, c2, kernel_size=3, padding=1, bias=False)
        
        # self.conv2 = DeformableConv2d(c2, c2, k, s, autopad(k, p), groups=g, bias=False)
        self.bn1 = nn.BatchNorm2d(c2)
        self.act1 = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
        self.bn2 = nn.BatchNorm2d(c2)
        self.act2 = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())

    def forward(self, x):
        # print(x.shape)
        # print('-'*50)
        x = self.act1(self.bn1(self.conv1(x)))
        # print(x.shape)
        offset = self.conv2_offset(x)
        x = self.act2(self.bn2(self.conv2(x,offset)))
        # print('-'*50)
        # print(x.shape)
        return x

修改yolo.p文件
找到parse_model函数，把DCNConv加入进去

 if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv,
                 BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x, CoordAtt, DCNConv):

修改yolov5s.yaml文件

# YOLOv5  by Ultralytics, GPL-3.0 license

# Parameters
nc: 1  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.50  # layer channel multiple
anchors:

	- [10,13, 16,30, 33,23]  
	- [30,61, 62,45, 59,119]  # P4/16
	- [116,90, 156,198, 373,326]  # P5/32

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, Conv, [64, 6, 2, 2]],  # 0-P1/2
   [-1, 1, DCNConv, [128, 3, 2]],  # 1-P2/4
   [-1, 3, C3, [128]],
   [-1, 1, DCNConv, [256, 3, 2]],  # 3-P3/8
   [-1, 6, C3, [256]],
   [-1, 1, DCNConv, [512, 3, 2]],  # 5-P4/16
   [-1, 9, C3, [512]],
   [-1, 1, DCNConv, [1024, 3, 2]],  # 7-P5/32
   [-1, 3, C3, [1024]],
   [-1, 1, SPPF, [1024, 5]],  # 9
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, Conv, [512, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 6], 1, Concat, [1]],  # cat backbone P4
   [-1, 3, C3, [512, False]],  # 13

   [-1, 1, Conv, [256, 1, 1]],
   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
   [[-1, 4], 1, Concat, [1]],  # cat backbone P3
   [-1, 3, C3, [256, False]],  # 17 (P3/8-small)

   [-1, 1, Conv, [256, 3, 2]],
   [[-1, 14], 1, Concat, [1]],  # cat head P4
   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)

   [-1, 1, Conv, [512, 3, 2]],
   [[-1, 10], 1, Concat, [1]],  # cat head P5
   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)

   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
  ]

运行下面命令查看网络结构

python models/yolo.py --cfg models/yolov5s.yaml

在自己的数据集上，map50提升了5%