yolov5s中加入DCNv2(可变形卷积v2)
Requirement
torch>=1.8.1
torchvision>=0.9.1
实现步骤
1.测试环境是否满足要求
import torch
import torchvision.ops
from torch import nn
import math
class DCNv2(nn.Module):
def __init__(self,
in_channels,
out_channels,
kernel_size=3,
stride=1,
padding=1):
super(DCNv2, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride if type(stride) == tuple else (stride, stride)
self.padding = padding
# init weight and bias
self.weight = nn.Parameter(torch.Tensor(out_channels, in_channels, kernel_size, kernel_size))
self.bias = nn.Parameter(torch.Tensor(out_channels))
# offset conv
self.conv_offset_mask = nn.Conv2d(in_channels,
3 * kernel_size * kernel_size,
kernel_size=kernel_size,
stride=stride,
padding=self.padding,
bias=True)
# init
self.reset_parameters()
self._init_weight()
def reset_parameters(self):
n = self.in_channels * (self.kernel_size**2)
stdv = 1. / math.sqrt(n)
self.weight.data.uniform_(-stdv, stdv)
self.bias.data.zero_()
def _init_weight(self):
# init offset_mask conv
nn.init.constant_(self.conv_offset_mask.weight, 0.)
nn.init.constant_(self.conv_offset_mask.bias, 0.)
def forward(self, x):
out = self.conv_offset_mask(x)
o1, o2, mask = torch.chunk(out, 3, dim=1)
offset = torch.cat((o1, o2), dim=1)
mask = torch.sigmoid(mask)
x = torchvision.ops.deform_conv2d(input=x,
offset=offset,
weight=self.weight,
bias=self.bias,
padding=self.padding,
mask=mask,
stride=self.stride)
return x
model = nn.Sequential(
DCNv2(3, 32, kernel_size=3, stride=1, padding=1),
nn.ReLU(inplace=True),
nn.MaxPool2d(2, 2),
DCNv2(32, 32, kernel_size=3, stride=1, padding=1),
DCNv2(32, 64, kernel_size=3, stride=1, padding=1),
nn.ReLU(inplace=True),
nn.MaxPool2d(2, 2),
DCNv2(64, 64, kernel_size=3, stride=1, padding=1),
DCNv2(64, 128, kernel_size=3, stride=1, padding=1),
nn.ReLU(inplace=True),
nn.MaxPool2d(2, 2),
DCNv2(128, 128, kernel_size=3, stride=1, padding=1),
DCNv2(128, 256, kernel_size=3, stride=1, padding=1),
nn.ReLU(inplace=True),
nn.MaxPool2d(2, 2)
)
x = torch.randn(2, 3, 64, 64)
y = model(x)
print(x.size())
print(y.size())
"""
torch.Size([2, 3, 64, 64])
torch.Size([2, 256, 4, 4])
"""
如果能输出,则说明环境适配。
2.修改models/yolov5s.yaml
# 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, DCNv2, [128, 3, 2]], # 1-P2/4
[-1, 3, C3, [128]], # 2
[-1, 1, DCNv2, [256, 3, 2]], # 3-P3/8
[-1, 6, C3, [256]], # 4
[-1, 1, DCNv2, [512, 3, 2]], # 5-P4/16
[-1, 9, C3, [512]], # 6
[-1, 1, DCNv2, [1024, 3, 2]], # 7-P5/32
[-1, 3, C3, [1024]], # 8
[-1, 1, SPPF, [1024, 5]], # 9
]
# YOLOv5 v6.0 head
head:
[[-1, 1, Conv, [512, 1, 1]], # 10
[-1, 1, nn.Upsample, [None, 2, 'nearest']], # 11
[[-1, 6], 1, Concat, [1]], # 12 cat backbone P4
[-1, 3, C3, [512, False]], # 13
[-1, 1, Conv, [256, 1, 1]], # 14
[-1, 1, nn.Upsample, [None, 2, 'nearest']], # 15
[[-1, 4], 1, Concat, [1]], # 16 cat backbone P3
[-1, 3, C3, [256, False]], # 17 (P3/8-small)
[-1, 1, Conv, [256, 3, 2]], # 18
[[-1, 14], 1, Concat, [1]], # 19 cat head P4
[-1, 3, C3, [512, False]], # 20 (P4/16-medium)
[-1, 1, Conv, [512, 3, 2]], # 21
[[-1, 10], 1, Concat, [1]], # 22 cat head P5
[-1, 3, C3, [1024, False]], # 23 (P5/32-large)
[[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]
3.修改models/common.py
# --------------------------DCNv2 start--------------------------
class DCNv2(nn.Module):
def __init__(self,
in_channels,
out_channels,
kernel_size=3,
stride=1,
padding=1):
super(DCNv2, self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride if type(stride) == tuple else (stride, stride)
self.padding = padding
# init weight and bias
self.weight = nn.Parameter(torch.Tensor(out_channels, in_channels, kernel_size, kernel_size))
self.bias = nn.Parameter(torch.Tensor(out_channels))
# offset conv
self.conv_offset_mask = nn.Conv2d(in_channels,
3 * kernel_size * kernel_size,
kernel_size=kernel_size,
stride=stride,
padding=self.padding,
bias=True)
# init
self.reset_parameters()
self._init_weight()
def reset_parameters(self):
n = self.in_channels * (self.kernel_size**2)
stdv = 1. / math.sqrt(n)
self.weight.data.uniform_(-stdv, stdv)
self.bias.data.zero_()
def _init_weight(self):
# init offset_mask conv
nn.init.constant_(self.conv_offset_mask.weight, 0.)
nn.init.constant_(self.conv_offset_mask.bias, 0.)
def forward(self, x):
out = self.conv_offset_mask(x)
o1, o2, mask = torch.chunk(out, 3, dim=1)
offset = torch.cat((o1, o2), dim=1)
mask = torch.sigmoid(mask)
x = torchvision.ops.deform_conv2d(input=x,
offset=offset,
weight=self.weight,
bias=self.bias,
padding=self.padding,
mask=mask,
stride=self.stride)
return x
# ---------------------------DCNv2 end---------------------------
4.修改models/yolo.py
if m in [Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv,BottleneckCSP, C3]:
#在列表加DCNv2
if m in [Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv,BottleneckCSP, C3, DCNv2]:
完成以上操作后,在train.py中导入对应的yaml文件和确认参数,即可开始训练。
参考
【1】 https://blog.csdn.net/shuaijieer/article/details/126249088
【2】 https://github.com/yjh0410/PyTorch_DCNv2