YOLOV5更换主干网络
GhostNet
更换GhostNet主干网络比较简单,因为models/common.py里面已经写好了相关类的定义,如下图
所以我们直接配置yaml文件即可
# YOLOv5 🚀 by Ultralytics, AGPL-3.0 license
# Parameters
nc: 7 # number of classes
depth_multiple: 0.33 # model depth multiple
width_multiple: 0.50 # layer channel multiple
anchors:
- [10, 13, 16, 30, 33, 23] # P3/8
- [30, 61, 62, 45, 59, 119] # P4/16
- [116, 90, 156, 198, 373, 326] # P5/32
# YOLOv5 v6.0 backbone
# ghostnet
backbone:
# [from, number, module, args]
[
[-1, 1, GhostConv, [64, 3, 2]], # 0-P1/2
[-1, 1, GhostConv, [128, 3, 2]], # 1-P2/4
[-1, 3, C3Ghost, [128]],
[-1, 1, GhostConv, [256, 3, 2]], # 3-P3/8
[-1, 6, C3Ghost, [256, 1]],
[-1, 1, GhostConv, [512, 3, 2]], # 5-P4/16
[-1, 9, C3Ghost, [512, 1]],
[-1, 1, GhostConv, [1024, 3, 2]], # 7-P5/32
[-1, 3, C3Ghost, [1024, 1]],
[-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)
]
并在yolo.py中进行一点修改:
最后在调用yaml文件时将原本的yoloXX.yaml切换为ghostnet.yaml即可。
parser.add_argument("--cfg", type=str, default="models/ghostnet.yaml", help="model.yaml path")
ShuffleNet
ShuffleNet的修改较为复杂,需要在models/common.py里面添加相关类的定义,我们直接写在common的末尾就可以:
def channel_shuffle(x, groups):
batchsize, num_channels, height, width = x.data.size()
channels_per_group = num_channels // groups
x = x.view(batchsize, groups, channels_per_group, height, width)
x = torch.transpose(x, 1, 2).contiguous()
x = x.view(batchsize, -1, height, width)
return x
class stem(nn.Module):
def __init__(self, c1, c2):
super(stem, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(c1, c2, kernel_size=3, stride=2, padding=1, bias=False),
nn.BatchNorm2d(c2),
nn.ReLU(inplace=True),
)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
def forward(self, x):
return self.maxpool(self.conv(x))
class Shuffle_Block(nn.Module):
def __init__(self, ch_in, ch_out, stride):
super(Shuffle_Block, self).__init__()
if not (1 <= stride <= 2):
raise ValueError('illegal stride value')
self.stride = stride
branch_features = ch_out // 2
assert (self.stride != 1) or (ch_in == branch_features << 1)
if self.stride > 1:
self.branch1 = nn.Sequential(
self.depthwise_conv(ch_in, ch_in, kernel_size=3, stride=self.stride, padding=1),
nn.BatchNorm2d(ch_in),
nn.Conv2d(ch_in, branch_features, kernel_size=1, stride=1, padding=0, bias=False),
nn.BatchNorm2d(branch_features),
nn.ReLU(inplace=True),
)
self.branch2 = nn.Sequential(
nn.Conv2d(ch_in if (self.stride > 1) else branch_features,
branch_features, kernel_size=1, stride=1, padding=0, bias=False),
nn.BatchNorm2d(branch_features),
nn.ReLU(inplace=True),
self.depthwise_conv(branch_features, branch_features, kernel_size=3, stride=self.stride, padding=1),
nn.BatchNorm2d(branch_features),
nn.Conv2d(branch_features, branch_features, kernel_size=1, stride=1, padding=0, bias=False),
nn.BatchNorm2d(branch_features),
nn.ReLU(inplace=True),
)
@staticmethod
def depthwise_conv(i, o, kernel_size, stride=1, padding=0, bias=False):
return nn.Conv2d(i, o, kernel_size, stride, padding, bias=bias, groups=i)
def forward(self, x):
if self.stride == 1:
x1, x2 = x.chunk(2, dim=1)
out = torch.cat((x1, self.branch2(x2)), dim=1)
else:
out = torch.cat((self.branch1(x), self.branch2(x)), dim=1)
out = channel_shuffle(out, 2)
return out
添加完成之后,我们再进行yaml文件的配置:
# YOLOv5 🚀 by Ultralytics, AGPL-3.0 license
# Parameters
nc: 7 # number of classes
depth_multiple: 0.33 # model depth multiple
width_multiple: 0.50 # layer channel multiple
anchors:
- [10, 13, 16, 30, 33, 23] # P3/8
- [30, 61, 62, 45, 59, 119] # P4/16
- [116, 90, 156, 198, 373, 326] # P5/32
# YOLOv5 v6.0 backbone
# shufflenet
backbone:
# [from, number, module, args]
[
[-1, 1, stem, [64]], # 0-P1/2
[-1, 1, Shuffle_Block, [128, 2]], # 1-P2/4
[-1, 3, Shuffle_Block, [128, 1]],
[-1, 1, Shuffle_Block, [256, 2]], # 3-P3/8
[-1, 6, Shuffle_Block, [256, 1]],
[-1, 1, Shuffle_Block, [512, 2]], # 5-P4/16
[-1, 9, Shuffle_Block, [512, 1]],
[-1, 1, Shuffle_Block, [1024, 2]], # 7-P5/32
[-1, 3, Shuffle_Block, [1024, 1]],
[-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)
]
然后在yolo.py中添加:
最后在调用yaml文件时将原本的yoloXX.yaml切换为shufflenet.yaml即可。
parser.add_argument("--cfg", type=str, default="models/shufflenet.yaml", help="model.yaml path")
MobileNet
MobileNet的添加我们采取另一种方式。
同样还是需要在models/common.py的末尾添加相关的类:
from torchvision import models
class MobileNetV3(nn.Module):
def __init__(self, n):
super().__init__()
model = models.mobilenet_v3_small(pretrained=True)
if n == 1:
self.model = model.features[:4]
elif n == 2:
self.model = model.features[4:9]
elif n == 3:
self.model = model.features[9:]
else:
print("其他")
def forward(self, x):
return self.model(x)
然后我们再配置yaml和修改yolo.py:
# YOLOv5 🚀 by Ultralytics, AGPL-3.0 license
# Parameters
nc: 7 # number of classes
depth_multiple: 0.33 # model depth multiple
width_multiple: 0.50 # layer channel multiple
anchors:
- [10,13, 16,30, 33,23] # P3/8
- [30,61, 62,45, 59,119] # P4/16
- [116,90, 156,198, 373,326] # P5/32
# YOLOv5 v6.0 backbone
# mobilenet
backbone:
# [from, number, module, args]
[
[-1, 1, MobileNetV3, [24, 1]], # 80 * 80
[-1, 1, MobileNetV3, [48, 2]], # 40 * 40
[-1, 1, MobileNetV3, [576, 3]] # 20 * 20
]
# YOLOv5 v6.0 head
head:
[[-1, 1, Conv, [512, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']], # 40 * 40
[[-1, 1], 1, Concat, [1]], # cat backbone P4 # 40 * 40
[-1, 3, C3, [512, False]], # 13
[-1, 1, Conv, [256, 1, 1]],
[-1, 1, nn.Upsample, [None, 2, 'nearest']], # 80 * 80
[[-1, 0], 1, Concat, [1]], # cat backbone P3 # 80 * 80
[-1, 3, C3, [256, False]], # 17 (P3/8-small)
[-1, 1, Conv, [256, 3, 2]], # 40 * 40
[[-1, 7], 1, Concat, [1]], # cat head P4 # 40 * 40
[-1, 3, C3, [512, False]], # 20 (P4/16-medium)
[-1, 1, Conv, [512, 3, 2]], # 20 * 20
[[-1, 3], 1, Concat, [1]], # cat head P5 # 20 * 20
[-1, 3, C3, [1024, False]], # 23 (P5/32-large)
[[10, 13, 16], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]
yolo的添加位置有所不同,需要添加俩个地方:
最后在调用yaml文件时将原本的yoloXX.yaml切换为mobilenet.yaml即可。
parser.add_argument("--cfg", type=str, default="models/mobilenet.yaml", help="model.yaml path")