前言：更改YOLOV5的backbone网络为 Shufflenetv2，便于达到轻量化的目的
1. 试运行YOLOv5
b站推土机
2. VOC数据集处理
3. 更改轻量级网络
参考魔改yolov5
3.1 在common.py末尾加入以下代码
#添加轻量化模块Shufflenetv2
# ---------------------------- ShuffleBlock start -------------------------------
# 通道重排，跨group信息交流
def channel_shuffle(x, groups):
batchsize, num_channels, height, width = x.data.size()
channels_per_group = num_channels // groups
# reshape
x = x.view(batchsize, groups,
channels_per_group, height, width)
x = torch.transpose(x, 1, 2).contiguous()
# flatten
x = x.view(batchsize, -1, height, width)
return x
class conv_bn_relu_maxpool(nn.Module):
def __init__(self, c1, c2): # ch_in, ch_out
super(conv_bn_relu_maxpool, 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, inp, oup, stride):
super(Shuffle_Block, self).__init__()
if not (1 <= stride <= 3):
raise ValueError('illegal stride value')
self.stride = stride
branch_features = oup // 2
assert (self.stride != 1) or (inp == branch_features << 1)
if self.stride > 1:
self.branch1 = nn.Sequential(
self.depthwise_conv(inp, inp, kernel_size=3, stride=self.stride, padding=1),
nn.BatchNorm2d(inp),
nn.Conv2d(inp, 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(inp 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) # 按照维度1进行split
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
# ---------------------------- ShuffleBlock end --------------------------------
3.2 在yolo.py parse_model里面添加新的模块
#新加模块conv_bn_relu_maxpool, Shuffle_Block
3.3 在model文件目录下新建yolov5.0s_shufflenetv2.yaml
# parameters
nc: 20 # number of classes
depth_multiple: 0.33 # model depth multiple
width_multiple: 0.50 # layer channel multiple
# anchors
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 backbone
backbone:
# [from, number, module, args]
# Shuffle_Block: [out, stride]
[[ -1, 1, conv_bn_relu_maxpool, [ 32 ] ], # 0-P2/4
[ -1, 1, Shuffle_Block, [ 128, 2 ] ], # 1-P3/8
[ -1, 3, Shuffle_Block, [ 128, 1 ] ], # 2
[ -1, 1, Shuffle_Block, [ 256, 2 ] ], # 3-P4/16
[ -1, 7, Shuffle_Block, [ 256, 1 ] ], # 4
[ -1, 1, Shuffle_Block, [ 512, 2 ] ], # 5-P5/32
[ -1, 3, Shuffle_Block, [ 512, 1 ] ], # 6
[ -1, 1, Conv, [ 1024, 3, 2 ] ], # 7-P5/32
[ -1, 1, SPP, [ 1024, [ 5, 9, 13 ] ] ],# 8
[ -1, 3, C3, [ 1024, False ] ], # 9
]
# YOLOv5 v5.0 head
head:
[[-1, 1, Conv, [512, 1, 1]], # 10
[-1, 1, nn.Upsample, [None, 2, 'nearest']],# 11
[[-1, 6], 1, Concat, [1]], # cat backbone P4 # 12
[-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]], # cat backbone P3 # 16
[-1, 3, C3, [256, False]], # 17 (P3/8-small)
[-1, 1, Conv, [256, 3, 2]], # 18
[[-1, 14], 1, Concat, [1]], # cat head P4 # 19
[-1, 3, C3, [512, False]], # 20(P4/16-medium)
[-1, 1, Conv, [512, 3, 2]],# 21
[[-1, 10], 1, Concat, [1]], # cat head P5
[-1, 3, C3, [1024, False]], # 23 (P5/32-large)
[[13, 17, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5)
]
3.4 更改train.py设置训练参数
4. 迁移学习训练改后网络
yolov5的freeze在train.py里面
freeze = [] # parameter names to freeze (full or partial)
freeze = ['model.%s.' % x for x in range(10,24)] # parameter names to freeze (full or partial)
for k, v in model.named_parameters():
v.requires_grad = True # train all layers
if any(x in k for x in freeze):
print('freezing %s' % k)
v.requires_grad = False
YOLOv5网络层的冻结通过设置其梯度为零来实现
本次冻结10到23层网络
freeze = ['model.%s.' % x for x in range(10,24)] # parameter names to freeze (full or partial)
for k, v in model.named_parameters():
v.requires_grad = True # train all layers
if any(x in k for x in freeze):
print('freezing %s' % k)
v.requires_grad = False
Class Images Labels P R mAP@.5 mAP@.5:.95: 100%|██████████| 155/155 [00:56<00:00, 2.73it/s]
all 4952 12032 0.683 0.607 0.648 0.394
aeroplane 4952 285 0.712 0.663 0.715 0.438
bicycle 4952 337 0.853 0.706 0.779 0.49
bird 4952 459 0.603 0.547 0.564 0.315
boat 4952 263 0.599 0.556 0.582 0.275
bottle 4952 469 0.524 0.301 0.322 0.156
bus 4952 213 0.783 0.657 0.736 0.578
car 4952 1201 0.777 0.749 0.805 0.549
cat 4952 358 0.734 0.703 0.754 0.513
chair 4952 756 0.5 0.374 0.375 0.187
cow 4952 244 0.675 0.52 0.588 0.335
diningtable 4952 206 0.703 0.544 0.624 0.404
dog 4952 489 0.672 0.578 0.646 0.398
horse 4952 348 0.742 0.747 0.776 0.47
motorbike 4952 325 0.821 0.734 0.804 0.49
person 4952 4528 0.774 0.687 0.764 0.409
pottedplant 4952 480 0.605 0.388 0.418 0.173
sheep 4952 242 0.515 0.677 0.603 0.37
sofa 4952 239 0.653 0.569 0.601 0.385
train 4952 282 0.753 0.766 0.821 0.517
tvmonitor 4952 308 0.672 0.672 0.69 0.42
100 epochs completed in 7.559 hours.
Optimizer stripped from runs\train\exp23\weights\last.pt, 12.3MB
Optimizer stripped from runs\train\exp23\weights\best.pt, 12.3MB
参考
魔改yolov5
