文章目录
概要
提示：这里可以添加技术概要
例如：
openAI 的 GPT 大模型的发展历程。
添加注意力机制流程
#添加方式一：将注意力机制添加到额外的一层
首先找一份注意力机制的代码，比如：ParNetAttention
powershell
import numpy as np
import torch
from torch import nn
from torch.nn import init
class ParNetAttention(nn.Module):
def __init__(self, channel=512):
super().__init__()
self.sse = nn.Sequential(
nn.AdaptiveAvgPool2d(1),
nn.Conv2d(channel, channel, kernel_size=1),
nn.Sigmoid()
)
self.conv1x1 = nn.Sequential(
nn.Conv2d(channel, channel, kernel_size=1),
nn.BatchNorm2d(channel)
)
self.conv3x3 = nn.Sequential(
nn.Conv2d(channel, channel, kernel_size=3, padding=1),
nn.BatchNorm2d(channel)
)
self.silu = nn.SiLU()
def forward(self, x):
b, c, _, _ = x.size()
x1 = self.conv1x1(x)
x2 = self.conv3x3(x)
x3 = self.sse(x) * x
y = self.silu(x1 + x2 + x3)
return y在ultralytics\nn\modules\下新建一份attention.py文件，将注意力机制代码放进去。
打开ultralytics\nn\tasks.py文件首先引入刚才新建的注意力机制代码：
powershell
from ultralytics.nn.modules.attention import ParNetAttention如果注意力机制代码是需要输入通道数的，那么在parse_model方法中加上这行代码：
powershell
# 有通道数的注意力机制放在这
elif m is (Zoom_cat,SSFF,ParNetAttention):
c2 = ch[f]
args = [c2, *args]如果注意力机制代码是不需要输入通道数的，可以不加这个。
最后更改yaml文件，将这个注意力机制加在你想加的地方
比如我加在SPPF层后边：
powershell
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f_attention, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f_attention, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2f_attention, [512, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2f_attention, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 9
- [-1, 1, ParNetAttention,[]] # 10加完之后这层注意力机制就是第10层，在Head中>=10层的全部+1
比如：
- [[15, 18, 21], 1, WorldDetect, [nc, 512, True]] # Detect(P3, P4, P5)
- 就要变成
- [[16, 19, 22], 1, WorldDetect, [nc, 512, True]] # Detect(P3, P4, P5)
添加方式二：将注意力机制添加到其中一层，不引入额外的层
比如想将注意力机制加到c2f中，打开ultralytics\nn\modules\block.py
首先将注意力机制代码导入到block.py中，复制一份c2f代码：
powershell
class C2f(nn.Module):
"""Faster Implementation of CSP Bottleneck with 2 convolutions."""
def __init__(self, c1, c2, n=1, shortcut=False, g=1, e=0.5):
"""Initialize CSP bottleneck layer with two convolutions with arguments ch_in, ch_out, number, shortcut, groups,
expansion.
"""
super().__init__()
self.c = int(c2 * e) # hidden channels
self.cv1 = Conv(c1, 2 * self.c, 1, 1)
self.cv2 = Conv((2 + n) * self.c, c2, 1) # optional act=FReLU(c2)
self.m = nn.ModuleList(Bottleneck(self.c, self.c, shortcut, g, k=((3, 3), (3, 3)), e=1.0) for _ in range(n))
def forward(self, x):
"""Forward pass through C2f layer."""
y = list(self.cv1(x).chunk(2, 1))
y.extend(m(y[-1]) for m in self.m)
return self.cv2(torch.cat(y, 1))
def forward_split(self, x):
"""Forward pass using split() instead of chunk()."""
y = list(self.cv1(x).split((self.c, self.c), 1))
y.extend(m(y[-1]) for m in self.m)
return self.cv2(torch.cat(y, 1))重命名为：
powershell
class C2f_attention(nn.Module):
"""Faster Implementation of CSP Bottleneck with 2 convolutions."""
def __init__(self, c1, c2, n=1, shortcut=False, g=1, e=0.5):
"""Initialize CSP bottleneck layer with two convolutions with arguments ch_in, ch_out, number, shortcut, groups,
expansion.
"""
super().__init__()
self.c = int(c2 * e) # hidden channels
self.cv1 = Conv(c1, 2 * self.c, 1, 1)
self.cv2 = Conv((2 + n) * self.c, c2, 1) # optional act=FReLU(c2)
self.m = nn.ModuleList(Bottleneck(self.c, self.c, shortcut, g, k=((3, 3), (3, 3)), e=1.0) for _ in range(n))
self.attention_AdditiveBlock = AdditiveBlock(c2)
def forward(self, x):
"""Forward pass through C2f layer."""
y = list(self.cv1(x).chunk(2, 1))
y.extend(m(y[-1]) for m in self.m)
return self.attention_AdditiveBlock(self.cv2(torch.cat(y, 1)))
def forward_split(self, x):
"""Forward pass using split() instead of chunk()."""
y = list(self.cv1(x).split((self.c, self.c), 1))
y.extend(m(y[-1]) for m in self.m)
return self.cv2(torch.cat(y, 1))然后需要在__init___函数中声明注意力机制函数，比如：self.attention_AdditiveBlock = AdditiveBlock(c2) 这里注意力机制如果需要参数的话就写上一层的输出通道数，作为本层的输入通道数，这里我想将注意力机制添加在cv2层后边，那么我的参数就是cv2等的输出通道数也就是c2，
在forward中对哪一层使用注意力机制就可以放在哪一层，比如
powershell
return self.attention_AdditiveBlock(self.cv2(torch.cat(y, 1)))注意力机制就添加完成了，最后需要把新的C2f_attention注册一下，
步骤如下：
首先在block.py中引入C2f_attention
powershell
__all__ = (
"DFL", "HGBlock", "HGStem", "SPP",
"SPPF", "C1", "C2", "C3",
"C2f", "C2fAttn", "ImagePoolingAttn",
"ContrastiveHead", "BNContrastiveHead",
"C3x", "C3TR", "C3Ghost",
"GhostBottleneck", "Bottleneck",
"BottleneckCSP", "Proto", "RepC3", "ResNetLayer",
"RepNCSPELAN4", "ELAN1", "ADown", "AConv",
"SPPELAN", "CBFuse", "CBLinear", "RepVGGDW",
"CIB", "C2fCIB", "Attention", "PSA", "SCDown",
# --------------------------------添加注意力机制
"C2f_attention",
)然后再__init__.py中添加
powershell
from .block import (
C1, C2, C3, C3TR, CIB, DFL,
ELAN1, PSA, SPP, SPPELAN, SPPF,
AConv, ADown, Attention, BNContrastiveHead,
Bottleneck, BottleneckCSP, C2f,
C2fAttn, C2fCIB, C3Ghost, C3x,
CBFuse, CBLinear, ContrastiveHead,
GhostBottleneck, HGBlock,
HGStem, ImagePoolingAttn,
Proto, RepC3, RepNCSPELAN4,
RepVGGDW, ResNetLayer, SCDown,
# ---------------------添加注意力机制-------------
C2f_attention,
)以及__all__ = ("C2f_attention")都要添加。
最后在task.py中引入C2f_attention：
三个位置：
（1）from ultralytics.nn.modules import (C2f_attention）
（2）if m in { Classify, Conv, ConvTranspose, GhostConv,.........C2f_attention
(3)if m in {BottleneckCSP, C1, C2, C2f, C2fAttn, C3, C3TR, C3Ghost, C3x, RepC3, C2fCIB,C2f_attention}: args.insert(2, n) # number of repeats n = 1
最后在yaml文件中将C2F层替换为C2f_attention即可。
powershell
backbone:
# [from, repeats, module, args]
- [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
- [-1, 3, C2f_attention, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f_attention, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2f_attention, [512, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2f_attention, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 9